MIXED-MEMBER ELECTORAL SYSTEMS IN CONSTITUTIONAL CONTEXT TAIWAN, JAPAN, and BEYOND EDITED BY NATHAN F. BATTO, CHI HUANG, ALEXANDER C. TAN, & GARY W. COX NEW COMPARATIVE POLITICS Mixed-Member Electoral Systems in Constitutional Context Reformers have promoted mixed-member electoral systems as the "best of both worlds." In this volume, internationally recognized political sci- entists evaluate the ways in which the introduction of a mixed-member electoral system affects the configuration of political parties. The con- tributors examine several political phenomena, including cabinet post allocation, nominations, preelectoral coalitions, split-ticket voting, and the size of party systems and faction systems. Significantly, they also consider various ways in which the constitutional system-especially whether the head of government is elected directly or indirectly-can modify the incentives created by the electoral system. Part I of the book provides an in-depth comparison of Taiwan and Japan, both of which moved from single nontransferable vote systems to mixed-member majoritarian systems. These cases demonstrate that the higher the payoffs of attaining the executive office and the greater degree of cross-district coordination required to win it, the stronger the incentives for elites to form and stay in the major parties. In such a context, a country will move rapidly toward a two-party system. In Part II, the contributors apply this theoretical logic to other countries with mixed-member systems and find that executive competition has the same effect on legislative electoral rules in countries as disparate as Thailand, the Philippines, New Zealand, Bolivia, and Russia. The findings presented here demonstrate that the success of elec- toral reform depends not only on the specification of new electoral rules per se but also on the political context-and especially the constitu- tional framework-within which such rules are embedded. NEW COMPARATIVE POLITICS Series Editor Michael Laver, New York University Editorial Board Ken Benoit, Trinity College, Dublin Gary Cox, Stanford University Simon Hix, London School of Economics John Huber, Columbia University Herbert Kitschelt, Duke University G. Bingham Powell, University of Rochester Kaare Strom, University of California, San Diego George Tsebelis, University of Michigan Leonard Wantchekon, Princeton University The New Comparative Politics series brings together cutting-edge work on social conflict, political economy, and institutional development. Whatever its substantive focus, each book in the series builds on solid theoretical foundations; uses rigorous empirical analysis; and deals with timely, politically relevant questions. Curbing Bailouts: Bank Crises and Democratic Accountability in Comparative Perspective Guillermo Rosas The Madisonian Turn: Political Parties and Parliamentary Democracy in Nordic Europe Edited by Torbjdrn Bergman and Kaare Strom Political Survival of Small Parties in Europe Jae-Jae Spoon Veto Power: Institutional Design in the European Union Jonathan B. Slapin Democracy, Dictatorship, and Term Limits Alexander Baturo Democracy, Electoral Systems, and Judicial Empowerment in Developing Countries Vineeta Yadav and Bumba Mukherjee The Latin American Voter: Pursuing Representation and Accountability in Challenging Contexts Edited by Ryan E. Carlin, Matthew M. Singer, and Elizabeth J. Zechmeister The European Union and the Rise of Regionalist Parties Seth K. Jolly Gender Quotas and Democratic Participation: Recruiting Candidates for Elective Offices in Germany Louise K. Davidson-Schmich Mixed-Member Electoral Systems in Constitutional Context: Taiwan, Japan, and Beyond Edited by Nathan F. Batto, Chi Huang, Alexander C. Tan, and Gary W. Cox MIXED-MEMBER ELECTORAL SYSTEMS IN CONSTITUTIONAL CONTEXT Taiwan, Japan, and Beyond Edited by Nathan F. Batto, Chi Huang, Alexander C. Tan, and Gary W. Cox University of Michigan Press Ann Arbor Copyright © by the University of Michigan 2016 All rights reserved This book may not be reproduced, in whole or in part, including illustrations, in any form (beyond that copying permitted by Sections 107 and 108 of the U.S. Copyright Law and except by reviewers for the public press), without written permission from the publisher. Published in the United States of America by the University of Michigan Press Manufactured in the United States of America O Printed on acid-free paper 2019 2018 2017 2016 4 3 2 1 A CIP catalog record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Names: Batto, Nathan E, editor. I Huang, Chi, 1953- editor. I Tan, Alexander C., editor. I Cox, Gary W., editor. Title: Mixed-member electoral systems in constitutional context : Taiwan, Japan, and beyond / edited by Nathan E Batto, Chi Huang, Alexander C. Tan, and Gary W. Cox Description: Ann Arbor : University of Michigan Press, 2016. I Series: New comparative politics 1 Includes bibliographical references and index. Identifiers: LCCN 20150386381 ISBN 9780472119738 (hardback) I ISBN 9780472121588 (ebook) Subjects: LCSH: Proportional representation-Taiwan. I Proportional representation- Japan. I Representative government and representation-Taiwan. I Representative government and representation-Japan. I Political parties-Taiwan. I Political parties-Japan. I Elections-Taiwan. I Elections-Japan. I Electoral systems I BISAC: POLITICAL SCIENCE / Political Process / Elections. Classification: LCC JQ1536 .M59 2016 I DDC 324.6-dc23 LC record available at http://lccn.loc.gov/2015038638 Acknowledgments The genesis of this volume lies in an attempt to answer a basic research question that surfaced in a roundtable of the second annual Asian Elec- tion Studies Conference held at the Election Study Center (ESC) of the National Chengchi University (NCCU), Taiwan, in May 2010. Why was it that Taiwan and Japan, two similar East Asian countries that had both switched from similar single nontransferable vote to similar mixed-member majoritarian legislative electoral systems, had displayed such divergent speeds and degrees in moving toward the theoretically expected consolida- tion of their party systems? A brainstorm with Gary Cox during his visit to the ESC in November 2011 sparked the idea of taking into account the constitutional context in which the electoral systems operate. As with any book-length project, we have a number of people and institutions to thank. We thank the ESC and its director, Lu-huei Chen, for continuous administrative support throughout the life of the project. A book with twenty-one coauthors based in fifteen institutions spanning four countries needs some sort of institutional base, and the ESC provided that home. We also wish to acknowledge our four home institutions, the Institute of Political Science at Academia Sinica, National Chengchi Uni- versity, the University of Canterbury, and Stanford University. In addition to paying our salaries, the four institutions have allowed us the time and resources necessary to complete this book. All the chapters are original contributions prepared specifically for this book. Earlier versions of most of the chapters in this volume were presented at the workshop on "Elections in Taiwan and Japan under the vi Acknowledgments Mixed-Member Majoritarian System" held on April 20, 2012, at ESC of NCCU in Taiwan. The workshop was funded by the Election and Citi- zens' Participation (ECP) Project, a Top University grant from the Min- istry of Education in Taiwan with Chi Huang as the principal investigator. We would like to extend thanks to all the scholars from Taiwan, Japan, the United States, and New Zealand who participated in the workshop and patiently worked with us on completing this edited book project. We express our deepest appreciation and admiration to Lu-huei Chen, Yin-lung Chou, Shih-chan Dai, Tzu-ping Liu, Hung-chung Wang, Chung- li Wu, and Misa Nishikawa for their intellectual contributions and toler- ance in the face of our unreasonable requests. Finally, we have more personal debts of gratitude. Nathan Batto thanks his wife, Lavai Fu-yi Yang, for her patience, encouragement, and inspira- tion. Chi Huang thanks Election and Citizens' Participation Project assis- tant ZeQui He for his help in compiling the manuscript for this edited volume during the review process. Contents List of Abbreviations ix Introduction: Legislature-Centric and Executive-Centric Theories of Party Systems and Faction Systems 1 Nathan F Batto and Gary W Cox PART I. THE CONSEQUENCES OF MMM ON POLITICAL COORDINATION IN TAIWAN AND JAPAN O N E The Consequences of MMM on Party Systems 25 Chi Huang, Ming-Feng Kuo, and Hans Stockton T W O The Consequences of Constitutional Systems on Party Systems 52 Jih-wen Lin T H R E E LDP Factions under SNTV and MA/M 73 Yoshiaki Kobayashi and Hiroki Tsukiyama F o U R Executive Competition, Electoral Rules, and Faction Systems in Taiwan 102 Nathan F Batto and Hsin-ta Huang FIVE Innovations in Candidate Selection Methods 135 Eric Chen-hua Yu, Kaori Shoji, and Nathan F Batto sIx Post Allocation, List Nominations, and Preelectoral Coalitions under MMM 165 Kuniaki Nemoto and Chia-hung Tsai Viii Contents SEVEN Split-Ticket Voting under MMIM 194 T Y Wang, Chang-chih Lin, and Yi-ching Hsiao PART II. COORDINATION IN MIXED-MEMBER SYSTEMS IN COMPARATIVE PERSPECTIVE E I G H T Thailand and the Philippines under MMM 229 Allen Hicken N I N E Political Consequences of New Zealand's MMP System in Comparative Perspective 247 Matthew S. Shugart and Alexander C. Tan T E N Presidents and Blank Votes in the Bolivian and Russian Mixed-Member Systems 278 Nathan F Batto, Henry A. Kim, and Natalia Matukhno Conclusions: Mixed-Member Systems Embedded within Constitutional Systems 300 Chi Huang Contributors 311 Index 317 List of Abbreviations General Terms CSM candidate selection method HC House of Councillors HR House of Representatives MM mixed-member MMIM mixed-member majoritarian MMP mixed-member proportional Nocai average effective number of electoral parties in each district NPR effective number of electoral parties in PR tier NPres effective number of presidential candidates NS effective number of parliamentary parties NSMD effective number of electoral parties in SMD tier NSNTV effective number of electoral parties in SNTV tier NV effective number of electoral parties PR proportional representation SMD single-member district SMP single-member plurality SNTV single nontransferable vote Political Parties in Taiwan DPP Democratic Progressive Party KMT Nationalist Party (Kuomintang) x List ofAbbreviations NP New Party NPSU Non-Partisan Solidarity Union PFP People First Party TSU Taiwan Solidarity Union Political Parties in Japan CGP Clean Government Party (Komeito) DPJ Democratic Party of Japan JCP Japanese Communist Party JRP Japan Restoration Party JSP Japan Socialist Party LDP Liberal Democratic Party PNP People's New Party (Nihon Shinto) SDP Social Democratic Party INTRODUCTION Legislature-Centric and Executive- Centric Theories of Party Systems and Faction Systems Nathan F Batto and Gary W Cox Two central tenets of the New Institutional paradigm are that institutions shape incentives and that how they do so depends on the specific context in which they are embedded. Building on these tenets, this book argues that electoral systems are embedded within constitutional systems and that whether the head of government is directly or indirectly elected affects how the legislative electoral system shapes politicians' incentives. Parties everywhere care about winning both legislative seats and executive offices. This book keeps both kinds of payoff in view and analyzes how constitu- tional strictures have mediated politicians' reactions to new mixed-member electoral rules in several countries. Mixed-Member Electoral Systems Mixed-member electoral systems have been touted as having the poten- tial to be "the electoral reform of the twenty-first century" (Shugart and Wattenberg 2001, 1). The basic idea behind such systems is simple. Some legislators are elected in single-member districts (SMDs) under plural- ity rule, while others are elected in multimember districts under some version of proportional representation (PR). Ideally, this offers citizens 2 Mixed-Member Electoral Systems in Constitutional Context the "best of both worlds," combining the direct ties between representa- tives and voters that characterize SMD systems with the proportional- ity that defines PR systems, and thereby promoting moderate and stable politics (Shugart and Wattenberg 2001, chap. 25). Politicians all over the world have found these promises enticing, and, in recent years, countries opting for new electoral systems have more often than not chosen from the mixed-member family. Currently 32 countries use some version of a mixed-member system, including such diverse cases as Mexico, Ukraine, Andorra, South Korea, Venezuela, Romania, Jordan, Hungary, Mongolia, Lithuania, and Germany.' In this book, we focus most heavily on two prominent East Asian examples of mixed-member reform: Taiwan and Japan. However, many of the arguments we make are general; and empirically the book considers several other relevant cases-including Thailand, the Philippines, New Zealand, Bolivia, and Russia. The angle from which we examine mixed- member reforms hinges on the distinction between legislature-centric and executive-centric theories of electoral systems, to which we turn next. Legislature-Centric and Executive-Centric Theories Standard theories of electoral systems often start by assuming that actors mainly want to win seats in the national legislature. Such theories then ana- lyze how different electoral systems push seat-maximizing parties toward different strategies in vote coordination, internal party organization, nomi- nations, collaboration with other parties, and so on. We call such theories legislature-centric because they focus on how the electoral rules governing pursuit of legislative seats affect strategies. In these theories, the key factor driving politicians is their recognition that groups smaller than a certain threshold, determined by the legislative electoral rules, will be unable to efficiently translate their popular support into seats. In other words, failure to coordinate invites punishment in the form of lost legislative seats. But of course politicians have other incentives to coordinate. In par- ticular, larger parties may sometimes be essential vehicles for the pursuit of executive office-both ministerial portfolios and the chief executive posi- tion (whether prime minister or president). Thus, another way to think about coordination, which one might call executive-centric, focuses on how efficiently groups of different sizes can convert their ambitions for execu- tive office into reality. In this line of theories, attention centers on the exec- utive electoral rules. While early contributions to executive-centric theory (e.g., Shugart 1995; Cox 1997, chaps. 10-11) focused on presidential elec- Introduction 3 tions, Hicken (2009) articulated a broader theory relevant to both nonpres- idential and presidential regimes. In the executive-centric line of theories, failure to coordinate invites punishment in the form of lost executive posts. In reality, politicians care about both legislative seats and executive offices. Their decisions to form larger parties or split into smaller ones should thus reflect their calculations about how such maneuvers will affect their payoff in both seats and offices, weighted by their concern for each. Because executive posts are generally viewed as more powerful and desir- able than legislative seats, one might even argue that theories of politi- cal coordination should be primarily executive-centric. This is consistent with work on party linkage or aggregation, such as Cox (1997, chaps. 10- 11), Chhibber and Kollman (2004) and Hicken (2009), which stresses the extent to which cross-district linkage of legislative candidates is influenced by the extent of executive power. How to Analyze the Electoral Reforms in Taiwan and Japan To illustrate how constitutional systems adjust the incentives set up by electoral systems, this book starts with a narrow substantive focus. In particular, it looks at one of the classic topics in the electoral systems literature-the number of competitors-in the context of Taiwan and Japan. It then gradually expands outward, both in the range of politi- cal phenomena covered and in geographical scope. This introduction focuses on the evolution of party systems and faction systems after elec- toral reform in Taiwan and Japan. Precis Taiwan and Japan used similar single nontransferable vote (SNTV) elec- toral systems prior to reform and moved to similar mixed-member majori- tarian (MA/FM) systems after it. If one considers the electoral system in isolation, the different outcomes in the two countries may seem puzzling. In each case, standard theories would predict that moving from SNTV to MMM should have induced greater coordination among competitors, reducing both the effective number of parties and the internal factionaliza- tion of the major parties. Yet, these very similar reforms have had notice- ably different effects. In Japan, the hoped-for reduction in the effective number of parties has been rather slow in coming, and the number of fac- tions has, if anything, actually increased. In Taiwan, in contrast, the party 4 Mixed-Member Electoral Systems in Constitutional Context system almost immediately fell into a two-party mold after the reform of 2005, while the reform has seemed to have very little impact on the num- ber of factions. Bringing the constitutional systems into the analysis provides a much richer and more satisfying explanation. The different ways in which execu- tive posts were allotted in the two countries-parliamentary Japan and semipresidential Taiwan-provided a much stronger impetus for party consolidation in the latter. Factions have also been affected in both coun- tries, though the most important changes have been in areas other than the number of factions. In the next several sections, we flesh out the argument just sketched. We begin by providing a slightly more detailed account of what the electoral reforms in Japan and Taiwan entailed. The Substance of the Electoral Reforms in Japan and Taiwan This section provides a quick sketch of the electoral rules before and after reform. The rules are reviewed in full detail in chapter 1. Japan and Taiwan both used SNTV systems to conduct their elections prior to reform, and they now both use MMNIM systems. Japan held its first MMM election in 1996 and Taiwan followed suit in 2008. There were some modest differences in the countries' prereform sys- tems. All seats in Japan were elected by SNTV, and most districts had three to five seats with an average of 3.96. Taiwan's districts were slightly larger on average with an average of 5.68 seats per district. More significantly, Taiwan also had a party list tier, which elected 22% of the total seats. How- ever, voters did not vote directly for the list tier; instead, party lists were elected on the basis of votes from the SNTV tier. In Japan's first MMM election, 300 seats were elected in SMDs while 200 were elected in the list tier. In subsequent elections, the list tier was reduced to 180 seats. Taiwan's MA/FM system featured 73 SMDs, 34 party list seats, and six seats reserved for indigenous peoples in two SNTV dis- tricts. Note that in both countries, SMD seats outnumber list seats by a considerable margin. Among the more important differences, Taiwan has a single national list where Japan has eleven regional lists, and Taiwan does not allow dual candidacies while Japan does. In sum, while there are some notable differences in the electoral rules, readers should not lose sight of the overriding similarities. In both coun- Introduction 5 tries, politicians and voters who had become accustomed to competing under SNTV rules had to adjust to a new set of incentives presented by new MMM rules. A Legislature-Centric View of Electoral Reform As noted above, a legislature-centric perspective draws attention to the incentives created by the electoral system used to elect the national legis- lature. Since both Taiwan and Japan changed their electoral systems from SNTV to MMM, new incentives should have reshaped party and faction systems in predictable ways. Parties The literature on electoral systems suggests that the change from SNTV to MMM in Japan and Taiwan should have reduced the number of political parties. Since SNTV has multiseat districts and voters can only support one candidate, a plurality is not necessary for victory. The more seats there are, the lower the threshold for victory is. Moreover, because parties do not need to win a plurality, there is no need to compete over the median voter. Small parties can build their support bases around nonmedian platforms (Cox 1990). SNTV should thus be able to sustain a multiparty system. In contrast, MMIM combines the logic of single-member-plurality (SMP) and list PR. Duvergerian incentives toward bipolar competition under SMP rules are well known (see Duverger 1954; Rae 1967; Cox 1997). Since a plurality is necessary for victory, smaller competitors tend to be weeded out, either because voters choose not to waste their votes on small parties or because, after suffering a string of defeats, small parties simply die off. If the two competitors are the same in all districts across the country, SMP should lead to a two-party system. In MMM systems, this Duvergerian logic is mitigated by the existence of the list tier. Small parties may not be able to win nominal seats, but they can win list seats. However, small parties do not get full proportional value for their votes. Because the two tiers are not linked, small parties win a proportional number of list tier seats, but this still leaves them under- represented in the SMDs and thus in the full chamber. Moreover, because the nominal tier is so much bigger than the list tier in East Asian systems, the proportionality of the list tier is overwhelmed by the majoritarian out- 6 Mixed-Member Electoral Systems in Constitutional Context comes of the nominal tier. The MA/IM systems in Japan and Taiwan thus put severe pressure on small parties, creating incentives for a shift toward a two-party system. As Chi Huang, Ming-Feng Kuo, and Hans Stockton document (chap- ter 1), these expectations have been met much more closely in Taiwan than in Japan. In Taiwan, the first MMM election saw the effective number of parties in the legislature plunge from 3.26 in 2004 to a mere 1.75 in 2008; this is exactly what conventional theories would predict. In Japan, how- ever, the trend has not been so clear. In fact, compared to the 1986 and 1990 SNTV elections, the first two MA/FM elections had a higher effective number of parliamentary parties. There has been a slow reduction of par- ties over the past decade, but this decrease in the number of parties has taken five elections under the new system. Japan's change has been slow and gradual; Taiwan's has been quick and dramatic. Factions A legislature-centric view suggests that electoral reform also should have affected the internal factions of the major parties in Japan and Taiwan. It is no simple task to define precisely what a faction is. One set of scholars has tried to conceptualize factions by looking at various critical dimensions, such as a group's organization, stability over time, ideological orientation, or its propensity toward rent-seeking.3 The problem is that factions often evolve and shift across these various lines. A second set of scholars worries less about the specific manifestation of the groups and more about the fact that a subgroup exists and contests power. Boucek, building on Zariski (1960) and Belloni and Beller (1978), defines factionalism as "the partition- ing of a political party ... into sub-units which are more or less institution- alized and who engage in collective action to achieve their members' par- ticular objectives" (2009, 14). One particular advantage of this definition for our project is that the locus of the competition determines the nature of the faction. For example, competition over control of local governments implies locally oriented factions (Belloni and Beller 1978). However, while Boucek's conceptualization is intentionally broad and inclusive, in this book we are only interested in the subset of factions that try to win power by nominating legislative candidates and securing executive posts.4 Factionalization of parties has been prominent in both Japan and Tai- wan, and several authors have argued that the SNTV electoral system promoted factionalism. Most accounts are based on the experience of the Japanese Liberal Democratic Party (LDP) and can be summarized as fol- Introduction 7 lows. A party wishing to win a majority of seats must nominate multiple candidates in most districts, and the competition within the party for nominations is intense. Factions arise as nominating coalitions, and they help their members run credible campaigns by centrally collecting finan- cial resources. Factions generally avoid supporting more than one person in each district, and the ideal situation for a faction would be to nominate exactly one candidate in each district (Thayer 1969, chap. 2; Fukui 1970, chap. 5; Curtis 1971, chap. 1; Baerwald 1986, 22; Curtis 1988, 82-86; Cox and Rosenbluth 1993).5 Perhaps the clearest exposition of a legislature-centric theory of fac- tions comes from Kohno (1992), who argues that there were five major factions during the SNTV era because the largest electoral districts had five seats. Supporting his view, Kohno shows that, while the four bigger factions nominated candidates in most districts, the smaller Komoto fac- tion nominated most of its candidates in five-seat districts.6 In Taiwan, Rigger (1999) tells a similar story about the Democratic Progressive Party's (DPP) factions in the 1990s. One of the factions' major roles is to help members win nominations. For much of the democratic era, DPP nominations were decided, at least in part, by allowing party mem- bers to vote. Factions recruit both candidates and party members, and they mobilize their members to vote on behalf of their candidates. District mag- nitude varies much more widely in Taiwan than in Japan, so factions do not always try to nominate their own candidate in each district. Instead, they often trade support across district lines. Money does not play as important a role in DPP as in LDP factional politics, but the DPP factions do provide electoral resources in the form of association with a national figure. Both in the nomination contest and in the general election, a DPP candidate can brand himself as "the Chen Shui-bian candidate," for example (Rigger 2001, chap. 5). If the SNTV electoral system was, in fact, the critical factor in sus- taining multiple intraparty factions, we might see changes after electoral reform. In MMM, each district has only one seat, so a party should only nominate one candidate. The literature suggests that competition for nominations is much like competition for seats, in that the decision rule is critical in determining how much consolidation to expect. If nominations are determined by a plurality rule, standard Duvergerian logic suggests that two big factions should form to contest each nomination. However, a different rule, such as requiring an absolute majority, may allow smaller factions some bargaining space in which to maneuver (Key 1984 [1949]; Duverger 1954; Canon 1978; Shugart and Carey 1992). 8 Mixed-Member Electoral Systems in Constitutional Context What is the decision rule for nominations in Taiwan and Japan? In Taiwan, contested nominations are now decided by telephone polls. The procedures are less transparent in Japan, where back-room negotiations determine most contentious questions (see chapter 5 for more on nomina- tions in Japan and Taiwan). Telephone polls use a plurality rule, and the single-shot contest does not allow runners-up any opportunity to use their demonstrated strength as bargaining chips. Japanese back-room negotia- tions allow smaller factions much more opportunity to trade their support in return for other concessions. Thus Taiwan's nominations can be consid- ered as using the plurality rule, while Japanese nominations can be seen as employing a majority rule. This implies that the pressures toward two factions should be stronger in Taiwan than in Japan. However, the pressure to consolidate the faction system in Japan should not be underestimated. The literature suggests that majoritarian systems will not support an unlimited number of actors. For example, a runoff sys- tem will theoretically only support three candidates (Cox 1997). Empiri- cally, majoritarian systems have, in fact, discriminated against small parties (Farrell 2011). In sum, there should be reductive pressure in both coun- tries, though it should be stronger in Taiwan than in Japan.7 To recap, both the LDP and DPP had four big factions and several smaller ones in the SNTV era. If the factions were, in fact, products of the electoral systems, the new MMM systems should have brought about some consolidation. In Taiwan, the new system should have pushed toward two big factions and perhaps a few smaller ones. In Japan, the pressures would have been weaker but still clearly reductive. This, however, is not a good description of the actual empirical record. To show this, we shall consider the LDP, DPP and Kuomintang (KMT) factions in turn. The five major LDP factions have continued to exist in the postre- form era. They have become less influential in the area of nominations (Cox, Rosenbluth, and Thies 1999, 42-47), but the continued presence of so many factions suggests that contesting nominations was not the only, or perhaps even the major, reason for their existence. In fact, the LDP faction system has actually expanded. In addition to the five old factions, four newer factions have emerged. Rather than consolidating into fewer factions, the LDP system has trended in the opposite direction. The DPP faction system has also undergone significant changes, but the current system is arguably just as complex as the prereform system. In the 1990s, the DPP had four big factions and several smaller ones. These factions were formally organized and physically headquartered in legisla- Introduction 9 tive offices. Today, factions are all informally organized, and only one of the old factions survives. Two other major factions are led by former premiers and presidential aspirants. A fourth faction centered on former President Chen Shui-bian is waning in strength, while a fifth faction led by the 2012 presidential candidate is growing and may soon rival the big three fac- tions. Several smaller factions also continue to operate. This reshuffling of the faction system is not compatible with legislature-centric expecta- tions. There were major changes in the early 2000s, long before electoral reform, when the factions aligned in the direction of two big coalitions. In 2006, after the electoral reform had been passed and one might expect a two-faction system to emerge, the system was reorganized into multiple informal factions. KMT internal factions have never fit the SNTV electoral model very well. At the national level, the KMT has had three periods of clear fac- tional divisions during the democratic era. In the early 1990s, the party was often divided into the mainstream and nonmainstream camps, led by President Lee Teng-hui and Premier Hau Pei-tsun, respectively (Tien 1996; Lin and Teddards 2003). In the late 1990s, the KMT was again split into two factions, with one led by Lee and his proteg6, Vice Presi- dent Lien Chan, and the other led by Governor James Soong (Diamond 2001). Finally, in the period before the 2008 election, the party was split into supporters of Speaker Wang Jin-pyng and former Taipei City mayor Ma Ying-jeou. All three of these splits can be understood as struggles between the KMT's nativist and orthodox Chinese wings. Two of the splits occurred during the SNTV era, while the third happened as the new electoral system was being drawn up, yet all three featured two big factions. District magnitude in the legislative electoral system seems to have made no difference. This impression is strengthened by looking at the KMT's local fac- tions.8 The KMT's local factions are based in a specific county. Since legis- lative districts in the SNTV era were counties, an electoral-systems theory of factionalism would predict that the number of local factions should depend on the number of seats available in each county. In fact, there were almost always two local factions. This was true in both small counties and big counties. For instance, in Taichung County, which had between six and 11 seats, the two KMT factions generally each nominated two or three leg- islative candidates. Moreover, local factions commonly nominated multiple candidates in each district for the County Assembly elections, which were also conducted under SNTV rules. The KMT's local factions have not seen any dramatic changes since electoral reform. 10 Mixed-Member Electoral Systems in Constitutional Context Adding an Executive-Centric Perspective We have reviewed legislature-centric arguments that the electoral reforms in Japan and Taiwan should have reduced the number of parties and fac- tions; and pointed out that the empirical record does not fit well with the theoretical predictions. In this section, we begin the process of explain- ing this discrepancy between theory and evidence by adding an executive- centric perspective. In this perspective, the pursuit of executive posts drives politicians, so the ways in which these posts are allotted is crucial. Parties From an executive-centric standpoint, the opening of executive positions to competition from opposition parties was an important milestone in the development of both countries' party systems. Both Japan and Taiwan experienced a long period of one-party dominance. In Taiwan, one-party dominance was legally mandated until 1986. Thereafter, it continued for some time because of the KMT's enormous resource advantages. In Japan, one-party dominance was not legally mandated. However, U.S. disapproval of the Japan Socialist Party (JSP) strongly militated against their participa- tion in government, as did the LDP's resource advantages as the party of government. During the period of one-party dominance in Taiwan, the opposition's incentive to coalesce was lessened by the common belief that they could not win, even were they better coordinated. Prior to 1986, the opposition had no chance of peacefully assuming power. From 1986 until 1991, the KMT's grip on the Legislative Yuan was assured by the presence of main- land representatives; and the opposition might have doubted the KMT's willingness to hold fair presidential elections or to relinquish power should they lose. Similarly, during the period of one-party dominance in Japan, the opposition faced a significant disincentive to coordination. U.S. oppo- sition to JSP participation in government both reduced the chance that an opposition alliance could win and reduced the value of winning. The struc- tural barriers to opposition governance in both countries meant that poor coordination was not the only, or main, reason the opposition believed executive power was unobtainable. Thus, the opposition could fragment without paying a penalty in lost executive offices, as those offices were already beyond its grasp for other reasons. The period of one-party dominance ended in the early 1990s in Japan and in the early 2000s in Taiwan. Once the prime ministership in Japan Introduction II and the presidency in Taiwan became realistic goals, opposition parties had greater incentives to settle their differences and present a unified front in the competition for the top spot and to make compromises necessary to bring new segments of voters into their coalition. From an executive- centric perspective, the key reforms in both Japan and Taiwan were those removing structural barriers impeding opposition competition for execu- tive office. In Taiwan, this meant lifting martial law and legalizing new political parties (1987), mandating the retirement of the mainland repre- sentatives (1991), and directly electing the entire legislature (1992) and the president (1996). In Japan, structural barriers lessened as U.S. opposition to JSP involvement in government relaxed, especially after the Cold War's end. In both cases, the period of one-party dominance-with its associ- ated and widespread belief that the opposition could never win power was dramatically shattered by a somewhat unexpected rotation of power. Both of these rotations came about because of splits in the old dominant party, not because the opposition first cobbled together a majority in the electorate. However, the triumph of the seven-party coalition in Japan in 1993 and the DPP's victory with a 39% plurality in the 2000 presidential election both made it clear that the executive was now in play. The Japanese and Taiwanese responded to the onset of competitive pol- itics for control of the executive quite differently. In Taiwan, the top prize, the presidency, could be won only by winning a plurality in the presidential election. Competition for the presidency thus drove the system toward two main candidates who, in turn, had strong incentives to organize legislators behind their candidacies. Theoretically, politicians may take some time to adapt to the coordi- nating incentives of plurality elections. Indeed, the recognition that the first two parties to establish themselves as viable routes to the presidency will likely be permanent features of the electoral landscape can raise the initial stakes and prevent coordination in the short run, as suggested by Fillipov, Ordeshook, and Shvetsova (1999). The best example of this on the Taiwanese scene is the formation of the People First Party after the 2000 election. After running a rebel campaign in 2000, thereby throwing the presidency to the DPP, James Soong found it easier to form a new party and contest the leadership of the Blue camp9 from outside, than to try to return to the KMT fold and secure its nomination. In the event, Soong was not able to establish himself as the consensus candidate and ended up accepting the second spot on a joint ticket in 2004. A merger of the two parties has been a continual possibility, especially after it became clear the People First Party would not displace the KMT as leader of the Blue camp 12 Mixed-Member Electoral Systems in Constitutional Context (Fell 2008). A somewhat different hiccup on the way toward bipartism was the Taiwan Solidarity Union, formed by former president Lee Teng-hui for the explicit purpose of leading his followers from the Blue camp into the Green camp (Wu 2002). All told, then, the emergence of plurality competition for the presi- dency was the main engine driving a bipolarization of politics into a Blue camp/Green camp format with echoes of the French quadrille bipolaire. The SNTV system for legislative elections provided space for multiple parties to survive within each camp. After the change to MAM, this space for small parties shrank dramatically, and the party system quickly reduced to the two presidential contestants. In Japan, things were different. Because the prime minister was elected indirectly, by the Diet, there were two viable routes to winning the prime ministership and other ministerial positions. One was to organize a large enough party (or coalition) to win a parliamentary majority at the legisla- tive elections and then divide the ministerial portfolios among the victors. Call this the electoral strategy. A second option was to win as many seats as possible and then, if no other party won a majority, bargain one's way into office after the election. Call this the postelectoral strategy. At this point, one can see why the emergence of close competition for the top executive position in Japan might not lead to bipartism in the Diet. Indirect election of a cabinet can be like electing a plural executive (M seats) using a two-stage electoral process. To the extent this analogy holds, one expects up to M+ 1 viable competitors. To put the point another way, if enough competitors for the prime ministership (and other executive offices) pursue a postelectoral strategy, then multiple parties can compete in the legislative elections, leaving the final selection to legislative bargain- ing. It is only if enough competitors for executive office opt for an electoral strategy that one expects a reduction in the number of parties (or the emer- gence of preelection alliances, on which see Carroll and Cox 2007). By this account, the emergence of real competition for the top execu- tive post(s), rather than electoral reforms, drove the evolution of the party system in both Japan and Taiwan. The responses differed because direct election of a powerful president demanded an electoral strategy, whereas indirect election of a relatively less powerful prime minister allowed a mix- ture of electoral and postelectoral strategies. Factions What accounts for the various changes or lack of changes in the parties' faction systems? An executive-centric explanation stresses the number of Introduction 13 important posts at stake. With a directly elected, nondivisible prize such as the presidency, competitors tended to divide into two factions. Where the prize was divisible, as with cabinet posts, it was not so important to aggre- gate into just two big factions. Multiple factions could survive by bargain- ing for a share of cabinet positions. An executive-centric perspective makes the persistence of multiple LDP factions more understandable. Most accounts of Japanese factions cite the electoral system as one reason for the factions' existence, but also stress the importance of bargaining over cabinet positions. Indeed, since the 1950s, the LDP's seniority system has ensured that cabinet positions are distributed more or less proportionally among the different factions. The factions have also fought for their fair share of other senior posts. While the electoral system changed, the parliamentary system and the number of important cabinet posts available did not. Since the number of cabinet posts is quite high, Japan's system could theoretically support even more factions. In fact, Yoshiaki Kobayashi and Hiroki Tsukiyama (chapter 3) suggest that the emergence of the four newer LDP factions is due, at least in part, to the prime minister's preference to give important posts to smaller rival factions. KMT factionalism also makes much more sense from an executive- centric perspective. All three periods of heightened factionalism at the national level were struggles over the presidency. In the latter two, the question was which faction would represent the KMT in the presidential elections of 2000 and 2008. In the first case, the question was whether the president or the premier would dominate politics in the new democratic system. Likewise, local factionalism was based on competition over execu- tive power. County executives have been directly elected since the 1950s, and many local factions can trace their history back to those first elections. Almost all accounts of KMT local factions stress winning the county gov- ernment and controlling its resources as the most important goal. Winning legislative or county assembly seats in SNTV elections is decidedly less important. The shifts in the number of DPP factions require a more complex explanation. Before 2000, DPP politicians focused on winning seats in the legislature, and a legislature-centric explanation is quite useful. In the late 1980s and early 1990s, with only a small number of seats open to elec- tions and not much support in the electorate, the DPP often could only nominate one candidates and it had two main factions, a more radical New Tide faction and a more moderate Formosa faction. As the number of seats up for election increased and the party's popularity grew in the 1990s, the DPP nominated more candidates per district. At the same time, 14 Mixed-Member Electoral Systems in Constitutional Context two new factions, the Welfare Alliance and the Justice Alliance, arose. In 2000, Chen Shui-bian won the presidency and the faction system changed dramatically. The Formosa faction disintegrated, Chen's Justice Alliance grew dramatically, and the Justice Alliance and New Tide faction forged a coalition, creating a clear distinction between those in power and those out of power. As Chen's two terms in office neared an end, the faction system changed once again, moving away from the nascent bifactional system. The DPP formally dissolved its factions in 2006, allowing the main presiden- tial contenders an opportunity to build new coalitions around themselves. However, the DPP has not seen the emergence of a pure bifactional sys- tem; in addition to the factions built to support President Chen and three presidential aspirants, one old faction survived and several smaller ones formed. The other factions seem to be operating on the legislature-centric model rather than focusing on winning the presidential nomination. They organize primarily to win posts in SNTV elections, such as those for the legislative party list, city councils, or the DPP's Central Executive Com- mittee, and they trade support across district lines to win nominations in single-seat elections, such as those for local mayors or seats in the legisla- tive nominal tier. Thus, DPP factions are currently organized around a mixture of executive-centric and legislature-centric logics. Combining the legislature-centric and executive-centric perspectives also provides insight toward another interesting question, why factions vary so much in their institutionalization. When factions are mainly vehi- cles to place their paramount leader in high executive office, they can be rather kaleidoscopic, rising and falling rapidly with their leaders' fortunes. In contrast, when factions become organizations to provide nomination and election assistance to all their legislative members, as well as helping them move up a lengthy hierarchy of internal posts, then they will be much more stable and institutionalized. Japanese and Taiwanese factions fit this argument quite well. LDP fac- tions help a legislator advance through every step of his or her career, and LDP factions are famously highly institutionalized. The main purpose of KMT factions has been to fill the top executive office, and KMT factions have been more loosely organized. The national factions faded away as soon as the contest over the presidency was resolved. Some local factions, such as the Taichung factions examined in chapter 4, have had long life spans, yet even they are informally organized. Other local factions are much more ephemeral and seemingly reorganize whenever a new person is elected into the top local office. Indeed, the in-power faction in many townships is referred to simply as the "town hall faction," and it is often opposed by Introduction 15 "the farmers' association faction." The DPP has experienced both sides of the spectrum. In the 1990s, when factions were primarily focused on win- ning seats in the legislature and controlling party posts, DPP factions were formally organized and quite stable. When the presidency came into play, the factions were formally abolished, factional affiliations became far more ambiguous, and factional activities become far less rigidly organized. Dual Candidacy An executive-centric perspective may also help us understand why Japan allowed dual candidacy while Taiwan did not. The LDP's factions had famously resisted electoral reform for decades. When they could no longer block reform, they still perceived multiple executive prizes for which they might compete. The current occupants of these plum posts faced a choice between adopting intraparty rules allowing dual nomination or not. Allowing dual candidacy (and operating them in the way the LDP did) improved the liquidity of the market in which the factional chiefs bar- gained for LDP nominations for their followers. Without dual candidacy, the nominations that the factions could trade came in four basic denomina- tions, ordered from most to least valuable: safe slots on the party list; safe districts; marginal slots on the party list; and marginal districts. Allowing dual candidacy created a fifth denomination, less valuable than any of the others: a marginal zombie slot.10 This allowed finer trades among the fac- tions, helping them avoid bargaining failures (which would take the form of multiple conservative candidates in the SMDs). It is not surprising that the LDP chose to implement party rules that allowed zombies, given that the Japanese factions had long experience in bargaining over the LDP nomination under the SNTV system; that the array of executive offices for which they competed had not changed; and that the cost of bargaining failure was higher under the new electoral system. In contrast to Japan, Taiwan had a much simpler bargaining context and did not adopt dual candidacy. When there was one preeminent prize, the presidency, that was reasonably attainable, Taiwanese parties tended to have two main factions or coalitions vying for that position. These factions had less need to create a small-denomination coin (zombie slots) in order to ensure that their bargaining over legislative nominations did not break down, simply because the bargain between two was simpler than the bar- gain between many. Moreover, the consequence of bargaining failure was less severe. Because the presidency was directly elected, losing a seat in the 16 Mixed-Member Electoral Systems in Constitutional Context legislature due to a nomination squabble would not immediately translate into a lower probability of winning the chief executive position, as it would in parliamentary Japan. Structure of the Volume The remainder of this volume will explore and develop the themes laid out in this introduction, starting from a narrow focus and moving outward. Part I compares the experiences of Japan and Taiwan. The first four chap- ters expand directly on the sketches of the party and faction systems laid out in the introduction. The next three chapters move to other aspects of the political systems in the two countries to show that many other aspects of politics can also be fruitfully explored by considering both a legislature- centric and executive-centric perspective. Part II moves beyond Japan and Taiwan to other countries that have mixed-member electoral systems. In countries as diverse as Thailand, the Philippines, New Zealand, Bolivia, and Russia, the way the executive is elected has reverberations throughout the political system, causing the mixed-member systems to operate quite differently in different countries. Chapter 1 lays out a concrete foundation for all the subsequent chap- ters. After describing the electoral rules in Taiwan and Japan in detail, Chi Huang, Ming-Feng Kuo, and Hans Stockton present the initial puzzle that following electoral reform, the number of parties decreased gradu- ally in Japan but very quickly in Taiwan. The authors then explore the degree to which this difference can be explained by the small variations in the electoral rules in the two countries, arguing that electoral rules can only account for part of the difference. Methodologically, Huang, Kuo, and Stockton present their argument in the framework of a most similar systems design and an interrupted time series with electoral reform as the interrupting event. This approach is adopted throughout part I of this vol- ume, though it is implicit rather than explicit in the other chapters. Chapter 2 takes up the unresolved puzzle from chapter 1. Since electoral rules cannot fully explain the numbers of parties, Jih-wen Lin examines the impact of the constitutional rules. Lin argues that Japan's parliamentary system provides opportunities for smaller parties to exercise influence over cabinet portfolios. In contrast, since cabinet positions in Taiwan's semi- presidential system are determined by the president, there is little incentive for politicians to split from the two big parties. Taken together, the first Introduction 17 two chapters provide a fuller explanation for the different party systems in Taiwan and Japan after electoral reform. Chapters 3 and 4 investigate changes in the faction systems in Japan and Taiwan. Following electoral reform in Japan, many analysts expected the LDP's factions to wither away and viewed their persistence as a bit of a puzzle. The most common explanation is that factions have retained an important role in allotting cabinet and party posts. Koshiaki Kobayashi and Hiroki Tsukiyama build on this notion by examining how the LDP party president has leveraged his institutional power to expand the influence of his own faction. Especially when the LDP president is personally popular, he has given his own faction a disproportionate share of nominations and posts. However, other factions remain viable because the prime minister still needs their support in parliamentary confidence votes. In Chapter 4, Nathan E Batto and Hsin-ta Huang discuss the limited impact electoral reform has had on the KMT's and DPP's internal factions. For both par- ties, competition over executive posts and the continuing use of SNTV in other elections has been more important to the faction systems than the adoption of MAM in legislative elections. Chapters 5, 6, and 7 move beyond a narrow focus on the numbers of competitors to look more broadly at how parties, candidates, and voters have adapted to the new electoral systems. In chapter 5, Eric Chen-hua Yu, Kaori Shoji, and Nathan E Batto examine candidate selection. In both Taiwan and Japan, parties have responded to the new electoral system by adapting new methods for identifying and nominating candidates. In Taiwan, polling primaries have become the default system for both major parties, while Japanese parties have increasingly turned to the kobo (public recruitment) system. Both of these institutions can be seen as attempts to select a candidate capable of appealing to pluralities or even majorities of voters in the district. Chapter 6 investigates trends in campaign strategies under the new electoral regimes. Kuniaki Nemoto and Chia-hung Tsai argue that the new electoral system has led to increasingly party-centered campaign strategies. Parties have increasingly used list nominations and cabinet positions strategically to try to enhance party popularity and fur- ther broad party goals. They have also reacted to the new system by form- ing preelectoral coalitions. However, the different institutional rules have affected the extent to which parties in Japan and Taiwan have adopted these new strategies. Chapter 7 looks at split-ticket voting. T. Y. Wang, Chang-chih Lin, and Yi-ching Hsiao argue that split-ticket voting takes place mostly within, rather than across, the two big camps. Supporters of 18 Mixed-Member Electoral Systems in Constitutional Context larger and smaller parties face quite different strategic contexts. In particu- lar, smaller parties have a strong incentive to educate their supporters on the different ways in which votes are counted in the two tiers, and this is reflected in patterns of split-ticket voting. Part II turns outward. While the theoretical logic presented in this intro- duction is grounded in the concrete realities of Japan and Taiwan, much of it can easily be transplanted to other contexts. In chapter 8, Hicken examines how the number of parties evolved in two Southeast Asian countries that have adopted MMM, Thailand and the Philippines. At first glance, these might appear to be counter-examples, since the presidential Philippines retains a fragmented party system while parliamentary Thailand has moved to a two-party system. However, Allen Hicken argues that these develop- ments are, in fact, quite compatible with the logic of executive competi- tion. In Thailand, reforms have produced highly nationalized politics and a dominant prime minister who has little need to compromise with other parties or minority factions within his party. In effect, what we have termed the postelectoral strategy was simply not viable in Thailand, and the result was, predictably, a two-party system. In the Philippines, Hicken argues that there are systemic disincentives for presidential hopefuls to engage in party-building. This results in a fragmented field in presidential elections, which, in turn, inflates the number of parties in the legislature. In chapter 9, Matthew S. Shugart and Alexander C. Tan move farther afield, looking at the mushrooming number of parties in New Zealand. Once considered the quintessential example of Westminster-style parliamentary democracy, New Zealand now has a quite fragmented party system, with as many as eight parties winning seats in parliament. Shugart and Tan argue that this fragmentation is due to both the permissive MMP electoral rules and the profitability of postelection strategies. The fragmentation has precluded majority party governments, and small parties have consistently been able to gain access to the cabinet. The result is a party system with two big par- ties and many small parties. In chapter 10, Nathan A. Batto, Henry E Kim, and Natalia Matukhno look at how presidential candidates affect voting in legislative elections in Bolivia and Russia, two countries very dissimilar to each other as well as to Taiwan and Japan. Voters in Bolivia and Russia have the option to vote "blank" or "against all," an option that is exercised far more often in the nominal tier than in the list tier. Many parties in these two countries are personal vehicles for presidential candidates and have difficulty fielding a full roster of credible district candidates. While voters might vote for the party list to support the presidential candidate, they are less likely to see connections between the larger party and the nominal Introduction 19 tier candidate, especially in inchoate parties. That is, they may not vote for the nominal tier candidate from their favorite presidential candidate's party, even if they are not attracted to any other district candidate. In fact, blank voting in the nominal tier is more common where inchoate parties are more popular. Finally, Chi Huang's conclusion brings the focus squarely back to the interactions between the constitutional system and the electoral system. In light of the theoretical and empirical lessons of this volume, Huang reminds us that similar institutions embedded within different concrete contexts can produce very different outcomes. NOTES 1. Data are from the International IDEA website. http://www.idea.int/esd/ (accessed January 29, 2014). 2. As is the standard practice, we use Laakso and Taagepera's (1979) index to operationalize our conceptual variable, coordination. We acknowledge that this index does not always accurately measure the degree of coordination in the sys- tem. For example, a value of roughly two can be obtained with two parties of .51 and .49 but also with three parties of size .66, .17, and .17. Several authors, such as Dunleavy and Boucek (2003) and Gaines and Taagepera (2013), have proposed new measures to address this deficiency. Nonetheless, the Laakso and Taagepera index has the advantages of simplicity and familiarity to most readers, so we continue to employ it while waiting for a consensus to emerge on a new standard. 3. For example, Duverger considers factions to simply be the product of how much internal diversity a party allowed its members (1954). Rose suggests that fac- tions must be organized and stable over time (1964). Sartori proposes that factions be classified on four dimensions, including organization, motivation, ideology, and their position on the Left-Right spectrum (1976). The International Comparative Political Parties project, led by Kenneth Janda, operationalized factions by measur- ing legislative cohesion, ideology, issues, leadership, strategy and tactics, and party purges (Janda 1993). For a review of different conceptualizations of factions, see Boucek (2009). 4. In privileging this subset, we follow in the tradition of V. O. Key, who consid- ers a faction to be any combination of actors who support a particular candidate (1984 [1949]). 5. Factions can also play a role in dividing the vote equally among a party's sev- eral candidates in multimember districts (see McCubbins and Rosenbluth 1995). 6. The Japanese literature pays less attention to the JSP factions but tends to view them also as products of the electoral incentives inherent in SNTV. 7. These pressures might be tempered somewhat by the list tier. Smaller fac- tions might survive by contesting list seats, though the relatively small list tiers in Japan and Taiwan limit the attractiveness of this survival strategy. 8. On the KMT's local factions, see Jacobs 2008; Rigger 1999, chap. 4; Chen 1996; Bosco 1992; and chapter 4, this volume. 20 Mixed-Member Electoral Systems in Constitutional Context 9. The Blue camp is a loose coalition of pro-unification parties, including the KMT, the PFP, and the New Party. The Green camp is a coalition of pro- independence parties, including the DPP and the TSU. 10. The Japanese media labeled dual candidates who lose in the district race but win a list seat as "zombies," since they have been killed in one tier but brought back to life in the other. REFERENCES Baerwald, Hans H. 1986. Party Politics in Japan. Winchester, MA: Allen & Unwin. Belloni, Frank P., and Dennis C. Beller, eds. 1978. Faction Politics: Political Parties and Factionalism in Comparative Perspectives. Santa Barbara, CA: ABC-Clio. Bosco, Joseph. 1992. "Taiwan Factions: Guanxi, Patronage, and the State in Local Politics." Ethnology 31:157-83. Boucek, Franeoise. 2009. "Rethinking Factionalism: Typologies, Intra-Party Dynamics and Three Faces of Factionalism." Party Politics 15 (4): 1-31. Canon, Bradley C. 1978. "Factionalism in the South: A Test of Theory and a Revisi- tation of V. O. Key." American Journal of Political Science 22 (4): 83 3-48. Carroll, Royce, and Gary W. Cox. 2007. "The Logic of Gamson's Law: Pre- election Coalitions and Portfolio Allocation." American Journal of Political Sci- ence 51:300-3 13. Chen, Ming-tong. 1996. "Local Factions and Elections in Taiwan's Democrati- zation." In Taiwan's Electoral Politics and Democratic Transition: Riding the Third Wave, ed. Hung-mao Tien. Armonk, NY: M. E. Sharpe. Chhibber, Pradeep, and Kenneth Kollman. 2004. The Formation of National Party Systems: Federalism and Party Competition in Canada, Great Britain, India, and the United States. Princeton: Princeton University Press. Cox, Gary W. 1990. "Centripetal and Centrifugal Incentives in Electoral Systems." American Journal of Political Science 34:903-35. Cox, Gary W. 1997. Making Votes Count. New York: Cambridge University Press. Cox, Gary W., and Frances M. Rosenbluth. 1993. "The Electoral Fortunes of Leg- islative Factions in Japan." American Political Science Review 87 (3): 577-89. Cox, Gary W., Frances M. Rosenbluth, and Michael P. Thies. 1999. "Electoral Rules, Career Ambition, and Party Structure: Comparing Factions in Japan's Upper and Lower Houses." American Journal of Political Science 44 (1): 115-22. Curtis, Gerald L. 1971. Election Campaigning Japanese Style. New York: Columbia University Press. Curtis, Gerald L. 1988. TheJapanese Way of Politics. New York: Columbia University Press. Diamond, Larry. 2001. "Anatomy of an Electoral Earthquake: How the KMT Lost and the DPP won the 2000 Presidential Election." In Taiwan's Presidential Pol- itics: Democratization and Cross-Strait Relations in the Twenty-First Century, ed. Muthiah Alagappa. Armonk, NY: M. E. Sharpe. Dunleavy, Patrick, and Franeoise Boucek. 2003. "Constructing the Number of Par- ties." Party Politics 9 (3): 291-315. Duverger, Maurice. 1954. Political Parties. New York: Wiley. Introduction 21 Farrell, David M. 2011. Electoral Systems: A Comparative Introduction. 2nd ed. New York: Palgrave Macmillan. Fell, Dafydd. 2008. "Inter-Party Competition in Taiwan: Toward a New Party Sys- tem?" In Presidential Politics in Taiwan: The Administration of Chen Shui-bian, ed. Steven M. Goldstein and Julian Chang. Norwalk, CT: Eastbridge. Fillipov, Mikhail G., Peter C. Ordeshook, and Olga V. Shvetsova. 1999. "Party Fragmentation and Presidential Elections in Post-Communist Democracies." Constitutional Political Economy 10:3-26. Fukui, Haruhiro. 1970. Party in Power: The Japanese Liberal-Democrats and Policy- Making. Berkeley: University of California Press. Gaines, Brian J., and Rein Taagepera. 2013. "How to Operationalize Two-Party- ness." Journal of Elections, Public Opinion, and Parties 2 3 (4): 3 87-404. Hicken, Allen. 2009. Building Party Systems in Developing Democracies. New York: Cambridge University Press. Jacobs, J. Bruce. 2008. Local Politics in Rural Taiwan under Dictatorship and Democracy. Norwalk, CT: Eastbridge. Janda, Kenneth. 1993. "Comparative Political Parties: Research and Theory." In Political Science: The State of the Discipline II, ed. Ada W. Finifter. Washington, DC: American Political Science Association. Key, V. O., Jr. 1984 [1949]. Southern Politics in State and Nation. Knoxville: Univer- sity of Tennessee Press. Kohno, Masaru. 1992. "Rational Foundations for the Organization of the Liberal Democratic Party in Japan." World Politics 44:369-97. Laakso, Markku, and Rein Taagepera. 1979. "Effective Number of Parties: A Mea- sure with Application to West Europe." Comparative Political Studies 12 (1): 3-27. Lin, Chia-lung, and Bo Teddards. 2003. "Lee Teng-hui: Transformational Leader- ship in Taiwan's Transition." In Sayonara to the Lee Teng-hui Era: Politics in Tai- wan, 1988-2000, ed. Wei-chin Lee and T. Y. Wang. Lanham, MD: University Press of America. McCubbins, Mathew D., and Frances Rosenbluth. 1995. "Party Provision for Per- sonal Politics: Dividing the Vote in Japan." In Structure and Policy in Japan and the United States, ed. Peter E Cowhey and Mathew D. McCubbins. Cambridge: Cambridge University Press. Rae, Douglas. 1967. The Political Consequences of Electoral Laws. New Haven: Yale University Press. Rigger, Shelley. 1999. Politics in Taiwan: Voting for Democracy. New York: Routledge. Rigger, Shelley. 2001. From Opposition to Power: Taiwan's Democratic Progressive Party. Boulder: Lynne Rienner. Rose, Richard. 1964. "Parties, Factions, and Tendencies in Britain." Political Studies 12:33-46. Sartori, Giovanni. 1976. Parties and Party Systems: A Framework for Analysis. New York: Cambridge University Press. Shugart, Matthew Soberg. 1995. "The Electoral Cycle and Institutional Sources of Divided Presidential Government."American Political Science Review 89 (2): 327-43. Shugart, Matthew Soberg, and John M. Carey. 1992. Presidents and Assemblies: Constitutional Design and Electoral Dynamics. New York: Cambridge University Press. 22 Mixed-Member Electoral Systems in Constitutional Context Shugart, Matthew Soberg, and Martin P. Wattenberg, eds. 2001. Mixed-Member Electoral Systems: The Best of Both Worlds? New York: Oxford University Press. Thayer, Nathaniel B. 1969. How the Conservatives Rule Japan. Princeton: Princeton University Press. Tien, Hung-mao. 1996. "Elections and Taiwan's Democratic Development." In Taiwan's Electoral Politics and Democratic Transition: Riding the Third Wave, ed. Hung-mao Tien. Armonk, NY: M. E. Sharpe. Wu, Yu-shan. 2002. "Taiwan in 2001: Stalemated on All Fronts." Asian Survey 42 (1): 29-38. Zariski, Raphael. 1960. "Party Factions and Comparative Politics: Some Prelimi- nary Observations." Midwest Journal of Political Science 4 (1): 27-51. PART ONE The Consequences of MMM on Political Coordination in Taiwan and Japan  ONE The Consequences of MMM on Party Systems Chi Huang, Ming-Feng Kuo, and Hans Stockton Electoral systems determine how votes cast in an election are translated into seats in the legislature, and thus to a large extent determine who wins and who loses in the political arena. The past two decades has seen a striking increase in the prevalence of mixed-member electoral systems. By combining the advantages of plurality rule and proportional represen- tation, these mixed systems attempt to strike a balance between the two. The intended and unintended consequences of such hybrid efforts have attracted considerable attention from scholars (e.g., Jou 2009; Ferrara 2004; Golder 2005; Huang 2011; Huang and Wang 2014; Huang, Wang, and Kuo 2008; Kostadinova 2002; Kohno 1997; Kuo, Huang, and Wang 2012; Lin 2008; Massicotte and Blais 1999; Moser and Scheiner 2004, 2012; Nishikawa and Herron 2004; Norris 2004; Reed 1999). Since the 1990s, democratic countries all over the world, as diverse as Italy, Scotland, Mexico, Bolivia, Venezuela, Russia, Albania, Hungary, Georgia, Lithuania, and New Zealand, have adopted mixed-member electoral systems. This is also true for many Asian countries, including Japan, Taiwan, Korea, Thai- land, and the Philippines. Despite their considerable differences in forms of government and political culture, congruent patterns in the new electoral systems are evi- dent across the Asian democracies. Japan, Taiwan, Korea, Thailand and the 25 26 Mixed-Member Electoral Systems in Constitutional Context Philippines all opted for mixed-member majoritarian (MMM) variants and all of them chose to weight the system heavily in favor of the majoritarian element and against the PR list (Reilly 2006, 2007). This chapter examines legislative elections in two Asian democracies, Taiwan and Japan, in order to answer a basic research question: why these two countries, having switched from single nontransferable vote (SNTV) to similar new MAM systems, displayed such divergent speeds and degrees of reaching theoretically expected political consequences. Specifically, we examine the change in party systems from multiparty to two-party systems. After electoral reform, Japan saw a gradual evolution from a multiparty to a two-party system over several election cycles, while the party system in Taiwan changed immediately and dramatically in the first postreform elec- tion (Huang 2011). Japan and Taiwan may constitute a no less comparable pair of subjects than Giannetti and Grofman's (2011) choice of Italy and Japan. A com- parative study of these two East Asian countries, while "controlling" for similarities, can shed some light on (1) the differential effects resulting from differing constitutional structure, and (2) how seemingly "minor" differences in electoral rules can disproportionately influence political consequences (see table 1.1 for a summary of electoral rules in Japan and Taiwan). We therefore adopt a quasi-experimental design by treating Japan and Taiwan as two nonequivalent but similar countries sharing the same sequence of electoral system change yet at different time periods a decade apart. When Japan experiences the change in 1996 Taiwan serves as a con- trol, and when Taiwan later experiences the change in 2008 Japan serves as the control. This design allows us not only to evaluate the speed and degree of the effects of electoral reform but to isolate potential causes of such effects as well. The remainder of this chapter is divided into five sections. The first section briefly describes the historical context of the formation of major parties in Japan and Taiwan, explains why the SNTV system helped to sustain the one-party dominant system, and summarizes other negative political consequences under SNTV. The second section traces the goals of electoral reform from SNTV to MMM systems in Japan and Taiwan. We then compare the rules of MMM systems in Japan and Taiwan by focus- ing on some subtle differences whose political consequences deserve closer examination. The third section presents detailed comparative analyses of these two countries with aggregate-level data, and furthermore we employ a piecewise regression of "interrupted time-series with switching replica- tions" quasi-experimental design to analyze their differences. The fourth The Consequences of MMM on Party Systems 27 section pinpoints some differences in similar electoral rules that partially solve the puzzle. The final section concludes that a complete solution requires a more general theory of electoral systems imbedded within con- stitutional structures, that is, the executive-centric theory that takes into account the political payoffs of the executive offices and the degree of elite coordination required to capture the executive offices. Party Politics under SNTV in Japan and Taiwan The evolutions of party systems in Japan and Taiwan have mirrored each other to a remarkable extent. Over time, these nascent democracies, each operating under dominant party arrangements, had transitioned into quite vibrant, multiparty systems. This was reversed as a consequence of elec- toral revisions that then led to the emergence of a two-party system under current circumstances. In Taiwan, the Nationalist Party (Kuomintang, KMT) has retained uninterrupted control of the country's legislature since the transition, although vacillating from single-party majority to majority in coalition and back to single-party majority. Japan's Liberal Democratic Party (LDP) experienced similar vacillation, albeit with a short interrup- tion after 1993 and again between 2009 and 2012. The LDP emerged in Japan in 1955 after the merger of two conserva- tive parties (the Liberals and the Democrats) that shared support among business interests and rural areas. The LDP dominated government for TABLE 1.1. Electoral Rules in Japan and Taiwan before and after Electoral Reform Japan Taiwan before after before after Total seats 511 500a 225 113 SNTV seats 511 176 6b Number of districts 129 31 2 Average seats per district 4.0 5.7 3.0 SMD seats 300 73 Party list seats 200a 49 34 Number of lists 11 2 1 Average seats per list 18.2a 24.5 34.0 Legal threshold for list none 5% 5% Dual candidacy allowed? yes no no Best loser provision? yes no no aIn 2000 and subsequent elections, the number of list tier seats and total seats were reduced to 180 and 480, respectively. The average number of seats per list declined to 16.4. bReserved for aboriginals. 28 Mixed-Member Electoral Systems in Constitutional Context 38 years until a stunning loss to a coalition of "clean government" parties in 1993 (Reed and Shimizu 2009, 5). The LDP quickly returned to power in 1996, but not until after the electoral reforms discussed in this chap- ter were pushed through. Despite the existence of robust opposition par- ties, such as the Japan Socialist Party, opposition parties were never quite strong enough to challenge the LDP's dominance. This party arrangement is commonly known as the 1955 System. Given the LDP's command of approximately two-thirds of the Diet seats from one election to the next, this 1955 System is also referred to as a One-and-a-Half Party System (Flanagan et al. 1991, 5; Scheiner 2006, 37).' Founded on the mainland in 1919, the KMT has matured during two distinct periods of time. The KMT was the ruling party of the Republic of China on the mainland for most of the period from 1919 to 1949. Since 1949, the KMT's jurisdiction has been limited to Taiwan and its surround- ing islands. Competitive party politics emerged after 1986, the transition to full democracy was completed in the early 1990s, and the longtime opposition Democratic Progressive Party (DPP) even won the presidency in 2000 and 2004. However, the KMT has retained continuous control of Taiwan's legislative branch. In the early part of the democratic era, the KMT retained a degree of dominance similar to that of the LDP. After 2001, splinter parties forced the KMT into coalition arrangements, but the party returned to a dominant status in the first postreform election of 2008 (see, for example, Stockton 2010). Both countries employed a SNTV electoral system for their national legislatures. In SNTV, each district has one or more seats and each voter can cast only one ballot for one specific candidate. There is no provision for preference rankings, so if a voter supports a candidate who does not win, the vote cannot be transferred to a second-favorite candidate. Simi- larly, if a candidate wins more votes than she needs, her excess votes cannot be transferred to other candidates. Japan's SNTV system was fairly straightforward. In 1993, the last elec- tion before reform, Japan elected 511 seats from 129 districts, an average of 3.96 seats per district. Each district had between two and six seats, and the great majority had three, four, or five seats. Taiwan's prereform system was more complex by comparison. In the 2004 election, there were a total of 225 seats. Of these, 176 seats were elected in 31 SNTV districts, for an average district magnitude of 5.68 seats per district.2 Several districts had only 1 seat, while the largest district had 13 seats. The median SNTV legislator was elected from a district with 8 seats. The Consequences of MMM on Party Systems 29 In addition to the 176 SNTV seats, there were also 49 seats elected by closed list PR on two separate lists. The list designated for national party representatives had 41 seats, while the list designated for overseas repre- sentatives had 8 seats. There was no separate party list ballot for the PR seats. Instead, all the votes for the party nominees running in the SNTV districts were summed to obtain each party's national total. For all parties with at least 5% of the national vote, these totals were used to apportion seats on the two lists using a largest remainders formula. Since there was no second ballot for the party list seats, only 22% of the seats were elected through the party lists, and each vote mattered so much more in the local race than in the national party totals,3 campaigns focused almost entirely on the SNTV portion of the election. As for processes of democratic transition, both Japan and Taiwan have experienced similar democratic development paths from a single dominant party system to a period of multiparty politics and then finally to a two- party system. Scholars have attributed one-party dominance as the reason for the implementation of SNTV electoral rules. Under SNTV, political parties must coordinate their supporters' votes within constituencies in order to more evenly distribute votes across candidates. Without success- ful coordination, weaker candidates will get too few votes while stronger copartisans will absorb too many votes. If effective, intraparty coordina- tion can result in political parties gaining an overrepresentation bonus. Attempting to benefit from the bonus of seats, a rational political party will nominate a bare number of candidates in each constituency in the SNTV electoral system if they can overcome the problems of coordina- tion of nomination and division of votes. Governing parties have a tremen- dous advantage because they can use the resources of the state to overcome these coordination problems (Cox 1996, 1997; Cox and Niou 1994; Cox and Rosenbluth 1993, 1996; Patterson and Stockton 2010; Rochon 1981). Many studies have found that a key reason for prolonged periods of LDP dominance was the LDP's ability to accomplish strategic nomination and vote coordination (Browne and Patterson 1999; Cox 1997; Horiuchi and Kohno 2004).4 If governing parties are expected to do well under SNTV, so are small parties. Small parties face much milder coordination problems since they often only nominate one candidate in any given district. With only one candidate, vote division is not a challenge (Taagepera and Shugart 1989, 28). Under SNTV, there is electoral space for small parties, and this should lead to a multiparty system. Several negative political consequences have been ascribed to SNTV. The vote share necessary to win a seat decreases as the number of seats 30 Mixed-Member Electoral Systems in Constitutional Context increases. For example, a candidate won a seat in a 16-seat district in Tai- wan in 1992 with only 2.6% of the total vote. Since candidates can win with support from a small minority of voters, they can appeal to nonmedian constituencies. Thus, candidates often took more extreme and sensational positions and appealed to highly personalized and niche voters (Flanagan et al. 1991). Correspondingly, Reed (1994) and Ramseyer and Rosenbluth (1997) also found that Diet members in Japan under the SNTV system were preoccupied with district service and fund-raising at the expense of their work in the legislature. Furthermore, given that party members are competing for both nominations and seats, SNTV accentuates intra- party competition, giving rise to factionalism and hampering interparty competition at the decisive election stage. In electoral campaigns, party platforms are of limited use since there are usually competing candidates running under the same party label. This means that campaigns are usually less focused on debates about national public policies. Instead, candidates have to emphasize personal qualities, such as local interests and social con- nections, and they typically build extensive and costly personal organiza- tions to mobilize voters. This localized, mobilization-based campaign style is much more expensive than one based on national debates over public policy, and the demand for funds inevitably leads to parties cozying up to corporations and systemic corruption (Cox 1997; Cox and Rosenbluth 1993; Yu Wang 2011). In summation, both Taiwan and Japan experienced numerous institu- tional challenges such as fractional politics, extremism, intraparty competi- tion, money politics, the inefficiency of parliament, and absence of power alternation during the development of democracy. Numerous studies traced these disadvantages to SNTV (Cox and Rosenbluth 1993; Cox and Thies 1998; Richardson 1988; Yeh-li Wang 2011). Because of continuous scandals and the inefficiency of parliament, subsequent electoral reform enjoyed widespread public support (Shiratori 1995; Yu Wang 2011). Electoral Reform and Differences in the MMM Systems Both Japan and Taiwan abandoned SNTV in favor of a new MNIM sys- tem. In both countries, proponents argued that electoral reform would produce a shift from the existing candidate-oriented campaign mode to policy-driven party politics and would ultimately improve government efficiency by encouraging alternations in government, such as in the adver- sarial two-party systems prevalent in the United States and the United The Consequences of MMM on Party Systems 3 1 Kingdom (Sakamoto 1999; Yeh-li Wang 2011). In the following section, we briefly describe the electoral reform, introduce the new electoral systems in Japan and Taiwan, and summarize their similarities and differences. In both countries the electoral reforms were initiated by a coalition/minority government trying to push for changes for its own sake, but the governing parties were eventually forced to compromise with the former dominant party maneuvering to stage a comeback. Although various electoral reform plans had been proposed for decades in Japan, it took defeat of the dominant LDP in the 1993 House of Rep- resentatives (HR) election to realize the reform. The outbreak of a series of corruption scandals in campaign finances, especially the Recruit scandal involving the Noburo Takeshita administration and Sagawa Kyubin scan- dal involving former LDP vice president Shin Kanemaru, paved the way for internal conflicts inside the LDP. Dissatisfied factions splintered off from the LDP and formed new parties such as the New Party Sakigake and the Japan Renewal Party. After the 1993 HR election, Morihiro Hosakawa took office by forming an eight-party anti-LDP coalition. The coalition government's bill of electoral reform passed the HR and yet was rejected by the House of Councillors in late 1993. Hosakawa chose to negotiate with the LDP and abandoned the coalition plan's single national PR constitu- ency by establishing eleven regional constituencies, a compromise with the LDP's prefectural PR constituencies. Finally, in January 1994 the Japanese Diet passed the electoral law reform bills, abolishing the old SNTV system and adopting a new MMIM system. This episode is recounted in detail by Curtis (1999) and Reed and Thies (2001). The 1994 electoral reform in Japan and its first implementation in 1996 sent a shockwave through neighboring Taiwan where the SNTV system had also been blamed for intraparty competition, factionalism, and money politics (see, for example, Cox 1996; Cox and Niou 1994). In the late 1996 National Development Meeting summoned by President Lee Teng-hui of the KMT, a consensus was reached to replace SNTV with a mixed-member system. However, a proposed constitutional amendment about electoral reform attempt failed in 1997 because the then-ruling KMT insisted on a Japanese-style MMIM system while the then-opposition DPP and New Party supported a German-style MMP system. Three years later, the DPP won the 2000 presidential election while the Pan-Blue parties maintained control of majority seats in the Legislative Yuan and dragged their feet over the DPP administration. After winning reelection in 2004, President Chen Shui-bian vigorously pushed for electoral reform by championing a proposal to "cut the assembly size in half' in order to "end the chaos in 32 Mixed-Member Electoral Systems in Constitutional Context the Legislative Yuan." KMT chair Lien Chan, worrying about public sup- port in the upcoming 2004 legislative election, threw his support behind the reform. The reform was unpopular with the rank-and-file legislators in both major parties, since they were not happy about half the seats in the legislature being eliminated. However, both caucuses went along with their party leaders and reluctantly voted in late 2004 to send the constitu- tional amendment proposal to the National Assembly for ratification. The National Assembly ratified the amendment in June 2005 with support from the two big parties. The smaller parties all opposed the reform, fearing a bleak future under the new MMM system. Mixed-member systems have two tiers of seats. In the nominal tier, vot- ers cast their votes for specific candidates. Most commonly, the nominal tier is conducted using single-member districts (SMDs) and the plurality rule, though there are other variants.5 In the list tier, seats are apportioned proportionally by party lists. Most mixed-member systems, including both the Japanese and Taiwanese variants, have a separate ballot for the list tier. In other words, voters cast two votes, one in the nominal tier to elect a rep- resentative from their local SMD, and one in the list tier to determine the apportionment of seats to the various party lists (Shugart and Wattenberg 2001,110-13).6 The Japanese system adopted in 1994 had 500 seats. In the nominal tier, 300 seats are elected by the plurality rule in SMDs. In the 1996 election, there were 200 seats in the list tier apportioned by the D'Hondt rule. In 2000 and in subsequent elections, the number of list tier seats was reduced to 180. The new Taiwanese system has a total of 113 seats. The nominal tier includes 73 seats elected by plurality in SMDs and six seats in two national SNTV districts for Mountain and Plains Aborigines. The remain- ing 34 seats comprise the list tier and are apportioned by a largest remain- der rule. Note that in both Japan and Taiwan, the nominal tier seats far outnumber the list tier seats. As such, the incentives created by the nominal tier may also be more powerful than those created by the list tier. In par- ticular, giving such weight to the nominal tier is damaging to smaller par- ties, since smaller parties often find winning a plurality in an SMD election to be a daunting challenge and rely heavily on seats from the list tier for survival. With fewer list tier seats available, smaller parties find it harder to survive. Furthermore, the drastic reduction in average district magni- tude undoubtedly increases the level of disproportionality and, usually, the effective number of parties. Despite the apparent similarities of the Japanese and Taiwanese MA/M systems, there are some subtle differences whose political consequences The Consequences of MMM on Party Systems 3 3 deserve closer examination. Here we concentrate on dual candidacy, the best loser provision, the proportion of seats allocated to the list tier, the number of PR constituencies, and the electoral threshold. The MAIM system in Japan allows dual candidacy, in which a candidate is allowed to register simultaneously in both the nominal and list tiers. If the candidate wins the SMD, his name is removed from the party list, and all the other candidates on the list move up one rank. A double-listed candidate who loses the SMD can still win a list seat. In popular Japanese parlance, candidates who are killed in the SMD races but come back to life in the list tier are known as zombies. While double candidacy is com- mon in mixed-member proportional systems around the world, in Japan this has proven to be one of the more controversial aspects of the MAM system. Table 1.2 shows that most parties in Japan other than the Clean Government Party (Komeito, CGP) and the Japan Communist Party take full advantage of this dual candidacy rule. In Taiwan, dual candidacy is not allowed, and seats are awarded by a strict closed-list rule. Japan further allows lists to be ordered on the basis of which dually nominated candidates prove to be the "best losers" in the nominal tier SMDs. Parties present lists in which multiple nominees are ranked at the same position. These ties are broken by results in the SMDs. Specifically, they are ranked by dividing their own SMD vote by the winning candi- date's vote. Losers who get a higher percentage of the winning total are ranked higher on the list. As with dual candidacy, parties make heavy use of the best loser provision, and this practice is increasing (see Nemoto and Tsai, chapter 6, this volume). Dual candidacy and the best loser provi- sion reduce conflict within the party by avoiding fights over rankings and provide an incentive for SMD candidates to win as many votes as possible even if they are unlikely to win the SMD seat (Reed and Thies 2001, 383). TABLE 1.2. Dual Candidacy Rates of Political Parties in Japan Japan Tomorrow Restoration Party of Japan LDP DPJ JCP SDP CGP Party (JRP) (TPJ) 1996 90.3 - 10.4 100.0 - - - 2000 95.9 98.8 11.3 100.0 38.9 - - 2003 92.1 98.9 10.3 100.0 0.0 - - 2005 96.6 98.6 8.0 94.7 0.0 - - 2009 93.1 98.9 39.5 100.0 0.0 - - 2012 96.2 100.0 4.0 100.0 0.0 100.0 98.2 Source: Japanese National Diet Library WARP Database, http://warp.ndl.go.jp/search/; http://www. soumu.go.jp/senkyo/senkyo_s/data/index.html. Asahi Shimbun Company, "Japan's Lower House Election Results, 2012" (CD-ROM) (Tokyo: Asahi Shimbun Company, 2013). 34 Mixed-Member Electoral Systems in Constitutional Context Japan allocates slightly more seats to the list tier than Taiwan. In Japan, 37.5% of seats are in the list tier, while only 30.1% of seats are elected by party lists in Taiwan. Taiwan has a single national list with 34 seats. To win any of these seats, parties must win at least 5% of the national list tier vote. This 5% legal threshold discriminates against smaller parties, since without a legal threshold parties winning at least 2.9% of the vote7 would be able to win a seat. In 2008, only the two big parties passed the threshold, and two small parties, the New Party (4.0%) and Taiwan Solidarity Union (TSU) (3.5%), were denied seats because of the threshold. In 2012, the TSU (9.0%) and the People First Party (5.5%) joined the two major parties in passing the threshold, and no other party got as much as 2.9%. In Japan, the 180 list tier seats are further distributed into 11 regional blocks, ranging from 6 to 29 seats in each PR constituency TABLE 1.3. Japan's PR Constituencies and Seats in HR Elections: 1996-2012 PR Blocs Areas Hokkaido Hokkaido Tohoku Amori, Iwate, Miyagi, Akita, Yamagata, Fukushima Kita-Kanto Ibaraki, Tochigi, Gunma, Saitama Minami- Chiba, Kanagawa, Yamanashi Kanto Tokyo Tokyo Hokuriku- Niigata, Nagano, Toyama, Shinetsu Ishikawa, Fukui Tokai Gifu, Aichi, Shizuoka, Mie Kinki Wakayama, Nara, Kyoto, Osaka, Shiga, Hyogo Chugoku Tottori, Shimane, Okayama, Hiroshima, Yamaguchi Shikoku Kagawa, Tokushima, Ehime, Kochi Kyushu Kagoshima, Fukuoka, Miyazaki, Saga, Oita, Kumamoto, Nagasaki, Okinawa Total 1996 2000 2003-12 9 8 8 16 14 14 Effective Thresholda 2003-12 8.33 5.00 21 23 20 20 3.57 21 22 3.26 19 17 17 4.17 13 11 11 6.25 23 33 21 30 21 29 3.41 2.50 13 11 11 6.25 7 6 6 10.71 23 21 21 3.41 200 180 180 4.31b Source: "Proportional Representation Districts and Seats of Japanese Lower House," Japanese Ministry of Internal Affairs and Communications website: http://www.soumu.go.jp/senkyo/senkyo_s/news/senkyo/ shu_teisu/index.html. Senkyoho Kenkyukai (Election Law Study Community), ed. Senkyoho to Seji shikinho tesatsu [The guide of Japanese electoral law and political funds control law]. Nagoya: Shinnippon-Hoki Publishing, 1995. aAccording to Lijphart (1997, 74), given the district magnitude m, the effective threshold is .75/(m + 1). bComputed based on the average number of seats: 16.4. The Consequences of MMM on Party Systems 3 5 (see table 1.3). Japan has no legal threshold, so the share of the list tier vote needed to win a seat is determined by the number of seats in each constituency. After Electoral Reform: Differences in the Political Consequences between Japan and Taiwan The differences in constitutional structures and electoral rules discussed above appear to have divergent political consequences in Japan and Taiwan, and they deserve careful and detailed comparative analyses. In this section, we detect and compare the consequences using electoral data based on macro-level patterns of party competition. After exploring the patterns of each country's party system formation, we advance a quasi-experimental research design referred to as a "switching replication" to test for any sub- stantial impact of electoral reform. Patterns of Party Competition Researchers often return to Duverger's classic arguments on electoral sys- tems and their political consequences. According to Duverger (1959, 217), "the simple-majority single ballot system favors the two-party system." He provides two factors to explain why a third party cannot survive in this electoral system. Duverger (1959, 224) argues that "the mechanical factor consists in the 'under-representation' of the third, i.e. the weakest party, its percentage of seats being inferior to its percentage of the poll." The second factor is a more ambiguous one. Because supporters of a third party do not want to waste their votes, Duverger (1959, 226) argues that it is "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." Duverger (1959, 205) thus asserted that single-member district plurality would tend to gener- ate two-party competition, and he also proposed that PR systems would encourage multiparty competition. Mixed-member electoral systems are characterized as hybrids of SMD and PR tiers. Since the primary feature of the MMM system adopted by Japan and Taiwan is the independent relationship between SMD and PR tiers, "the typical majoritarian boost received by a large party in the nomi- nal tier is not likely to be wiped away by proportional allocation from the list tier" (Shugart and Wattenberg 2001, 13). Furthermore, the PR portion in Japan accounts for only 3 7.5% of the 480 total seats since 2000 and in Taiwan accounts for only 30.1% of the 113 total seats. It seems logical to 36 Mixed-Member Electoral Systems in Constitutional Context argue that the gravity force of Duverger's law will exert pressure on small parties and thus push down the number of parties. Indeed macro-level data in tables 1.4 and 1.5 seem to conform to Duverger's law and indicate that the impact of electoral reforms on party systems in the two countries is a movement toward two-party competition, albeit at different speeds.8 Tables 1.4 and 1.5 present the Laakso-Taagepera effective number of electoral parties (N) and effective number of parliamentary parties (N) from an SNTV to a MMM system for these two countries (Laakso and Taagepera 1979; Taagepera and Shugart 1989). According to Duverg- er's law, the number of parties in SMDs would shift toward two due to mechanical effects and strategic voting due to psychological effects. But Duverger's hypothesis predicts that multiple parties remain in the PR tier of the system because voters have stronger incentives to vote sincerely. This is indeed what we see in table 1.4 for Japan's HR elections over time. The effective number of electoral parties in the SMD tier (NSMD) fluctuated between the first two elections after the 1994 reform as Japan TABLE 1.4. Effective Number of Parties in Japan: 1958-2012 Year NSNTV NSMD NPR Ns 1958 2.24 1.98 1960 2.40 2.00 1963 2.55 2.15 1967 3.03 2.41 1969 3.35 2.50 1972 3.41 2.67 1976 4.02 3.18 1979 3.76 3.29 1980 3.44 2.74 1983 3.64 3.23 1986 3.35 2.57 1990 3.42 2.70 1993 5.15 4.14 1996 3.86 4.28 2.93 2000 3.77 5.15 3.16 2003 2.97 3.42 2.59 2005 2.71 3.72 2.26 2009 2.64 3.66 2.10 2012 3.81 5.79 2.45 Source: Japanese National Diet Library WARP Database: http:// warp.ndl.go.jp/search/; Tokifumi Mizusaki and Mori Hiroki, Sousen- kyu no tokuhyou bunseki: 1958-2005 [Japanese general elections: 1958- 2005] (Tokyo: Bokutakusha, 2007), 26-27; Asahi Shimbun Company, "Japan's Lower House Election Results, 2012" [CD-ROM] (Tokyo: Asahi Shimbun Company, 2013). The Consequences of MMM on Party Systems 37 went through a period of party realignment but then dropped sharply below 3.0 parties for the three elections in 2003, 2005, and 2009.9 The effective number of electoral parties in PR tier (NPR) has also declined over time, though it has consistently been higher than its SMD coun- terpart and remained above 3.5 after the 2003 election. With more than 62% of the HR seats allocated to the SMD tier, it is not surprising that the NS dropped below the NSMD. In short, throughout the period from 1996 to 2009 in Japan, partisan politics were largely focused on the con- test between the Liberal Democratic Party (LDP) and the second largest party (e.g. the New Frontier Party in 1996 and the Democratic Party of Japan since 2000). As Jou (2009, 2010) points out, during this decade, party competition in Japan's SMD tier witnessed two distinct patterns: urban areas converging toward a two-party system while rural constituencies remaining dominated by the LDP (see also Yu Wang 2011). The landslide victory of the Demo- cratic Party of Japan (DPJ) with 64% of the seats in the 2009 HR elec- tion and the formation of the DPJ-centered coalition government appear to have ushered in a new era of government with the alternation of two major parties (Arase 2010; Krauss and Pekkanen 2011; Maeda 2010). The pendulum swinging back to the LDP with 61.3% of the total seats in the 2012 HR election is further evidence of this trend. But it also seems that MMM can produce extreme one-party dominance and decimate opposi- tion parties. The data for Taiwan's legislative elections are presented in table 1.5. These clearly indicate that there was an immediate and dramatic drop in all three indicators of the effective number of parties. Whereas the NSNTV TABLE 1.5. Effective Number of Parties in Taiwan: 1992-2012 Year NSNTV NSMD NPR NS 1992 2.52 2.46 1995 2.90 2.63 1998 3.14 2.53 2001 4.14 3.24 2004 3.76 3.08 2008 2.29 2.49 1.47 2012 2.32 3.03 1.97 Source: Data for 1992-2008 elections results from Election Study Center, National Chengchi University, website: http://vote.nccu. edu.tw/cec/vote4.asp?passl=B; data for the 2012 election result from Taiwan's Political Geography Information System (TPGIS) website: http://tpgis.nccu.edu.tw/MainPage/. 38 Mixed-Member Electoral Systems in Constitutional Context in the 2004 SNTV elections had been 3.76, in 2008 NSMD plunged to 2.29, just as Duverger's Law suggests. Furthermore, the NPR measure was a mere 2.49. It took Japan a decade and a half and five HR elections to evolve into a true two-party system, but this transformation was realized almost immediately in Taiwan. As in Japan, the dominance of the nominal tier led to a lower number of parliamentary parties than electoral parties. In 2008, the NS was a mere 1.47, which is much closer to one-party dominance than one would expect in a democracy. Even with a much stronger performance by the DPP and after the People First Party split from the Pan-Blue coali- tion and ran its own candidates in the 2012 legislative election, the NS rose only slightly to 1.97 and NSMD grew only to 2.32 and NPR 3.03, respectively. Evaluating the Impact of Electoral Reform: A Quasi-Experimental Design This study applies the logic of "most similar systems" design (MSSD) in the comparative politics tradition (Przeworski and Teune 1970) to the "interrupted time series" (ITS) quasi-experiment frequently used in the fields of policy evaluation (Campbell and Stanley 1963). Most similar sys- tems design involves matching two cases that experience different out- comes while appearing to be very similar in as many respects as possible except on key theoretical points. The goal of this design is to identify the difference that is responsible for contrasting outcomes. Those similarities between two cases can be considered relatively "controlled for" and thus ruled out as possible causes. The remaining differences between the two cases emerge as the candidates causing their divergent outcomes. MSSD thus facilitates causal inference. Japan and Taiwan are selected following the logic of the most simi- lar systems design. In other words, they are chosen for this study because they are relatively similar in many respects. Both are densely populated island countries along the Asian continent without rich natural resources and yet have built relatively healthy open economies with broadly distrib- uted wealth. Both have Confucian and Buddhist cultural heritages as well. Although Japan is under a parliamentary system and Taiwan is under a semipresidential system, both countries otherwise have similar governing institutions led by a longtime dominant clientelistic political party and dis- ciplined bureaucrats. Indeed, Taiwan was a Japanese colony for 50 years, and many of its institutional structures and behavioral habits can be traced directly to Japanese sources. The ITS design, on the other hand, involves one experimental group The Consequences of MMM on Party Systems 39 and repeated observations before and after an intervention (Shadish, Cook, and Campbell 2002, 175). The weakness of this single time series design is the threat of history to the internal validity of causality. That is, there is always the possibility that factors other than the event under investiga- tion come to influence the dependent variable at about the same time the intervention occurs. If we can find two nonequivalent but relatively similar countries, each of which has the same sequence of electoral system change yet at differ- ent times, then a certain degree of control can still be achieved. That is, when one country experiences intervention the other serves as a control, and when the control nation later experiences the event the original treat- ment case serves as the control (Shadish, Cook, and Campbell 2002, 192). Applying this "ITS with switching replications" design to this study, it can be diagrammed as: oooooXooooo ooooooooXoo where each 0 represents an observation of the dependent variable (i.e., NS), and X stands for the occurrence to the event (i.e., shift from SNTV to MMM system). Figure 1.1 displays the evolution of the effective number of parliamen- tary parties in Japan and Taiwan, respectively. In general, both countries experience a rising trend during the SNTV period and then witness a decline after adopting the MA/FM system. However, the decline is more abrupt and immediate in Taiwan than in Japan. In terms of a cross-country comparison, when Japan first implemented the MMM system in 1996 and started to show a slow decline trend in NS, Taiwan showed no such change. Similarly, when Taiwan first implemented the MA/M system and experi- enced a sharp decline in NS in 2008,Japan's NS had already reached a steady state of nearly two. This switching pattern strengthens our confidence in making a causal inference from the impact of electoral system change. Impact Analysis Box and Tiao's (1975) ITS analysis is often used to assess the impact of a discrete intervention on a social process (see, for example, Box,Jenkins, and Reinsel 2008; McCleary and Hay 1980). However, this method requires one to identify and estimate a noise model based on the preintervention 40 Mixed-Member Electoral Systems in Constitutional Context 45 4 2.5 Electoral reform bills Constitutional amendment 0.5 passed in Japan, 1994 in laiwan, 2005 0 Year - -- Japan's Ns --Taiwan's Ns Fig. 1.1. Effective number of parliamentary parties (NS): Japan and Taiwan (Note: Years with "*" denote Legislative Yuan elections in Taiwan, while years without "*" re- fer to lower house elections in Japan; "**" denotes general elections held in both Taiwan and Japan. Data from Japanese National Diet Library WARP Database: http://warp.ndl.go p/search/; Tokifumi Mizusaki and Mori Hiroki, Sousenkyu no tokuhyou bunseki: 1958-2005 [Japanese general elections: 1958-2005], 26-27 (Tokyo: Bokutakusha, 2007); Yomiuri Online: http:// www.yomiuri.co.jp/election/shugiin/2012/kaihyou/, accessed on March 28, 2013.) series as a benchmark for comparison with the entire series. Given the limitation of the frequency of legislative elections held in the past half cen- tury and thus the number of observations available for this study, the five (for Taiwan) to thirteen (for Japan) observations before electoral system changes occurred are obviously too few to apply Box and Tiao's techniques. We therefore turn to the piecewise linear regression approach of ITS (Greene 2012; Lewis-Beck 1986; Marsh and Cormier 2002), which takes advantage of the entire series and is still able to assess the impact of events. The method is to regress the dependent variable upon independent vari- ables representing a trend and events. Each event, in turn, is represented by a dummy variable (which captures the drop or jump in intercept change) and a postintervention time counter (which captures the slope change). Based on the piecewise linear regression approach, the equation we use to evaluate the impact of shifting from the SNTV to MMM system in Japan is specified as follows: Y = + PT+ fl, L+ PC,+ PD, + PC, + ,(1.1) where The Consequences of MMM on Party Systems 41 Y = dependent variables, NS T,= a counter for time from 1 to the last observation D = a dummy variable indicating the implementation of the MMM system in Japan, that is, D1t = 1 since year 1996 Ct= a postintervention time counter scored 0 before 1996 and 1, 2, 3 ... for observations on and after 1996 D = a dummy variable indicating the implementation of the MMM system in Taiwan, that is, D2t = 1 since year 2008 Ct= a postintervention time counter scored 0 before 2008 and 1, 2, 3 ... for observations on and after 2008. The equation for assessing electoral system change in Taiwan is similar to the equation (1.1) for Japan except that its observations start from 1992 when the island held its first free and open Legislative Yuan elections and thus consists of only seven observations up to the 2012 election. Given this extremely short time series and thus small degree of freedom, statistical estimates serve only as a supplement to the visual inspection of figures. If our hypothesis is correct, that is, if electoral system change causes a gradual shift in the party system in Japan while leading to an immediate align- ment in Taiwan, then we expect divergent patterns of impact in the two countries. That is, we expect no abrupt change in intercept D1t but only a slow change in slope C1t in Japan. On the other hand, we expect at least an abrupt change in intercept D2t and perhaps also in slope C2t in the case of Taiwan. Statistical results for the NS of Japan and Taiwan are presented in table 1.6. They generally confirm the visual inspection of figure 1.1. For both countries, the evolution of the NS changes only when each country shifts from the SNTV to the new MMM system. Furthermore, the general pat- tern of change fits our expectation. There is no significant change in slope since 1996 but only a slight decline in slope (Ct = -0.120) in the case of Japan. In contrast, there is an abrupt drop in level (D2t = -2.15 5) in the case of Taiwan in 2008,10 albeit no change in slope is detected perhaps due to the fact that only two elections have been held after the electoral reform. The ITS design uses Taiwan's reform as a control for Japan and Japan's reform as a control for Taiwan. In the Japan model, the variables for Taiwanese reform are not significantly different from zero, and a parallel result holds in the Taiwan model. This provides an added level of confi- dence that the changes in the party systems were, in fact, due to the treat- ment variable, electoral reform. Overall, what both figure 1.1 and table 42 Mixed-Member Electoral Systems in Constitutional Context 1.6 show is a gradual shift from a multiparty system to a two-party system taking place over several elections after reform in Japan, but a sudden and abrupt shift in the first postreform election in Taiwan. Why Different Speeds toward a Two-Party System? The differences between Japan and Taiwan discussed above pose an inter- esting question. Why did it take Japan almost one and a half decades to gradually consolidate toward a two-party system and strategic voting pat- tern (see Scheiner and Tronconi 2011, 102) while in Taiwan the impact of the new electoral system was immediate and extreme? To what extent are these different rates of change the result of different electoral rules, such as dual candidacy, the best loser provision, the number of party list con- stituencies, the threshold for list constituencies, and the proportion of seats allocated to the nominal and list tiers? We speculate that all these elements contribute to the divergent results. One reason why the number of electoral parties may not have declined as precipitously as expected has to do with contamination effects. A num- ber of studies have noted that in mixed-member systems, in order to lift the PR votes, party elites may field candidates in single-member districts to give the party label a human face, regardless of the chance of winning the TABLE 1.6. Impact of MMM on Ns: Japan and Taiwan Effective number of parliamentary Japan Taiwan parties (NS) Coefficient (S.E.) Coefficient (S.E.) Time Counter T 0.042*** (0.010) 0.057 (0.167) Adopting MMM System in Japan -0.367 (0.440) -0.2 30 (0.648) (D1t = 1 since 1996) Trend under MMM in Japan (Clt time counter since 1996) Adopting MMM System in Taiwan (D2 = 1 since 2008) Trend under MMM in Taiwan (C2t time counter since 2008) Constant R2 Adjusted R2 Number of observations -0.120t (0.056) 0.035 (0.187) -0.400 (0.777) -2.155# (0.707) 0.195 (0.186) 0.033 (0.151) 1.978*** (0.202) 2.403 (0.488) 0.651** 0.516 19 0.944 0.663 7 tp<0.1; p < 0.05; p <0.01; t p < 0.001. 'I t I = 3.05 but is insignificant (p > 0.1) due to too few observations and degrees of freedom in Taiwan. The Consequences of MMM on Party Systems 43 district seat. Electoral incentives in one tier thus "contaminate" those in the other tier, complicating the insights from Duvergerian laws (Cox and Schoppa 2002; Herron and Nishikawa 2001; Moser and Scheiner 2004). This might account for higher than expected N. However, these extra candidates are assumed to be sure losers, and they should not affect Ns. Moreover, contamination effects yield no insight into the differences in the evolution of the party system in Japan and Taiwan. Taiwan's list tier has a legal 5% threshold. In Japan, the average regional constituency has an implied threshold of 5.8%. This is marginally higher than in Taiwan, so one might conclude that the Japanese system is more unfriendly to small parties. However, because the number of seats on each regional list varies, so does the threshold. If we look at Lijphart's (1994, 27; 1997, 74) estimate for effective threshold," small parties in Japan have even more opportunities. In Shikoku, with only six seats, the effective threshold is 10.71% and is a very high barrier for small parties. In contrast, Kinki has 29 seats, and small parties might not be intimidated at all by the relatively low 2.5% threshold since a party with 2.5% has a reasonable chance to win a seat. In fact, six out of the 11 regional constituencies have thresholds under 5 % (see the last column of table 1.3). In short, there are opportuni- ties for small parties to win list seats in the larger regions, and small parties in Japan can strategically concentrate their efforts on regions in which they have a reasonable chance of exceeding the threshold. The differences in the thresholds imply that small parties should find more opportunities to win list seats in Japan than in Taiwan. Unlike in Tai- wan, they do not have to cross a relatively high 5 % national threshold to survive. Instead, they can win one or two seats with a somewhat lower vote and build on this foundation in future elections. They can also concentrate their resources in the most favorable regions and ignore less promising areas. These opportunities are further enhanced by the fact that Japan allo- cates a slightly higher percentage of seats to the list tier. We suspect that dual candidacy and the best loser provision also help small parties survive in Japan. Candidates who "die gloriously" in their SMDs can be "revived" in the PR tier as "zombies" and thus have strong incentives to cultivate local connections and campaign hard in their SMDs. This hard work, in turn, can result in higher party votes in the list tier (though see Maeda 2008 for an opposing view). Kuo, Huang, and Wang (2012) analyzed Japanese House elections from 1996 to 2009 and found evidence that smaller parties such as the Social Democratic Party and los- ing major parties such as the LDP in 2009 relied more on dual candi- dacy's lifting effects on PR votes. The Social Democratic Party did use 44 Mixed-Member Electoral Systems in Constitutional Context the dual candidacy strategy to improve their PR votes and furthermore to sustain their seats in parliament. In Taiwan, on the other hand, there is no mechanism for such interactions between SMD and PR ballots since no dual candidacy is allowed. Huang (2010) evaluated the effects of the TSU's nomination of 13 district candidates on their PR vote shares in the 2008 elections in Taiwan and indeed found no evidence of interaction between the SMD and PR tiers. Neither Taiwan's nor Japan's electoral system is friendly to small par- ties, but small parties have a better chance of survival in Japan's system. Yet we suspect that these contrasting electoral rules cannot fully explain why small parties have fared so much better in Japan than in Taiwan. A more complete explanation requires consideration of differences in the constitu- tional designs of each country, a topic Lin takes up in chapter 2 and others develop in the rest of this volume. Concluding Remarks Institutions do matter, and even seemingly "minor" differences may pro- duce significantly different outcomes. This chapter explores the effect on party systems of several such differences between Japan and Taiwan, including dual candidacy, the best loser provision, the proportion of seats allocated to the list tier, the number of PR constituencies, and the elec- toral threshold. Whether and to what extent these differences produce the divergent consequences discussed above call for further careful compara- tive studies and rigorous causal analyses. Japan and Taiwan undertook similar electoral reforms from SNTV to MMM, and they both experienced a similar development from a multi- party system to a two-party system. However, the speeds at which the party systems transformed after electoral reform varied significantly. We argue that the combination of some seemingly "minor" differences in electoral systems contributed to these divergent results by providing more oppor- tunities for smaller parties to survive in Japan than in Taiwan. However, we do not argue that differences in the electoral rules can fully account for the larger numbers of parties in Japan. Electoral systems cannot be delinked from the wider constitutional arrangements in which they occur. As Lin argues in the next chapter, analyses of party system change are best served when we focus not only on changing rules but also on the constitu- tional context in which they are embedded. Specifically, researchers should The Consequences of MMM on Party Systems 45 keep both legislative seats and executive offices in view while analyzing the effects of legislative electoral systems. NOTES 1. In the 1955 System, the main debates on national policies between the ruling LDP and opposition parties were mainly based on the conservative-progressive political ideology spectrum. Their debate was as follows: (1) whether Japan should move toward becoming a capitalist or a socialist state; (2) whether or not the ninth statement of Japan's constitution should be amended; (3) whether or not the U.S.-Japan Treaty (Ampou) alliance should be maintained; (4) in terms of defense, whether military forces in Japan should be enhanced or remain neutral (Yang 2002, 65). 2. In all, 168 representatives were elected from 29 geographically defined dis- tricts. Another eight members were elected from two nationwide districts reserved for Plains Aborigines and Mountain Aborigines. 3. Usually, 40,000 votes were sufficient to elect a district candidate, whereas each seat on the national party representatives list needed about 200,000 votes. 4. Some studies further indicates that in "1955 system," because the LDP usu- ally took two-thirds of the seats in the Diet, and the opposition parties could not cooperate with each other to challenge the LDP-dominant regime, the power alternation was only between the main factions inside the LDP, not alternation between parties (Reed 2003, 21). 5. The prereform Taiwanese system was technically a MMM system since it had a nominal tier, with voters choosing specific candidates in the SNTV tier, as well as a list tier. However, for the purpose of clarity, this volume will refer to the prereform system as an SNTV system. MMNM will refer narrowly to the postreform systems in Japan and Taiwan with nominal tiers composed of SMDs. 6. Mixed-member systems are commonly divided into two categories, mixed- member majoritarian (MMM) and mixed-member proportional (MMP). The criti- cal factor is whether the two tiers are linked. In MAM , the tiers are not linked, so seats are determined independently in each tier. That is, if a party wins 50% of the party list votes, it wins 50% of the list tier seats, regardless of whether it won all, some, or none of the nominal tier seats. The plurality formula used in the nominal tier can lead to significant disproportionality, and the list tier in MMM systems merely mitigates rather than erases this disproportionality. In MMP systems, the two tiers are linked in order to produce a proportional outcome. Each party's total seat share is determined by the list tier vote, and, depending on the number of seats it wins in the nominal tier, it is awarded the appropriate number of party list seats necessary to obtain a proportional share of seats in the overall chamber (Shugart and Wattenberg 2001, 13-17). 7. The theoretical threshold is 1/(m + 1). 8. By "speeds" we mean the number of elections it takes to reach similar results. 9. The NSMD remained high (3.77) in Japan's 2000 House election. This was partly because the coalition partnership of the LDP and CGP did not effectively 46 Mixed-Member Electoral Systems in Constitutional Context coordinate their nominations. Specifically, the LDP and CGP have formed a coali- tion to nominate candidates to compete against non-LDP candidates in all elec- tions since 2000. In most districts, CGP supporters were supposed to vote for an LDP candidate, and the LDP gave organizational votes to support CGP's PR list candidates. But in the 2000 election, the coalition did not work very well, as the LDP and CGP simultaneously fielded SMD candidates in four districts. Since the 2003 election, this problem has been overcome. See also Nemoto and Tsai, chapter 6, this volume. 10. With five variables and only seven cases in the Taiwan model, even large t-values do not reach conventional levels of significance. In the following pages we adopt a rule of thumb of a t-ratio above 2 as a rough guideline for interpreting the case of Taiwan. 11. The effective threshold is estimated as .75/(m + 1). REFERENC ES Arase, David. 2010. "Japan in 2009: A Historic Election Year." Asian Survey 50 (1): 40-55. Asahi Shimbun Company. 2013. "Japan's Lower House Election Results, 2012" [in Japanese] (CD-ROM). Tokyo: Asahi Shimbun. Box, George E. P., Gyilym M. Jenkins, and Gregory C. Reinsel. 2008. Time Series Analysis: Forecasting and Control. Hoboken, NJ: John Wiley and Sons. Box, G. E. P., and G. C. Tiao. 1975. "Intervention Analysis with Applications to Economic and Environmental Problem." Journal of the American StatisticalAsso- ciation 70 (349): 70-79. Browne, Eric C., and Dennis Patterson. 1999. "An Empirical Theory of National Nominating Behavior." British Journal of Political Science 29 (2): 259-89. Campbell, Donald T., and Julian C. Stanley. 1963. Experimental and Quasi- Experimental Designs for Research. Boston: Houghton Mifflin. Cox, Gary W. 1996. "Is the Single Non-Transferable Vote Superproportional? Evidence from Japan and Taiwan." American Journal of Political Science 40 (3): 740-55. Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World's Electoral Systems. Cambridge: Cambridge University Press. Cox, Gary W., and Emerson Niou. 1994. "Seat Bonuses under the Single Non- Transferable Vote System: Evidence from Japan and Taiwan." Comparative Poli- tics 26 (2): 221-36. Cox, Gary W., and Frances Rosenbluth. 1993. "The Electoral Fortunes of Legisla- tive Factions in Japan." American Political Science Review 87 (3): 577-89. Cox, Gary W., and Frances Rosenbluth. 1996. "Factional Competition for the Party Endorsement." British Journal of Political Science 26 (2): 2 59-69. Cox, Gary W., and Michael E Thies. 1998. "The Cost of Intraparty Competition: The Single, Non-Transferable Vote and Money Politics in Japan." Comparative Political Studies 31 (3): 267-91. Cox, Karen E., and Leonard J. Schoppa. 2002. "Interaction Effects in Mixed- Member Electoral Systems: Theory and Evidence from Germany, Japan, and Italy." Comparative Political Studies 35 (9): 1027-53. The Consequences of MMM on Party Systems 47 Curtis, Gerald L. 1999. The Logic ofapanese Politics. New York: Columbia Univer- sity Press. Duverger, Maurice. 1959. Political Parties: Their Organization andActivity in the Mod- ern State. 2nd English ed. New York: John Wiley and Sons. Farrell, David M. 2011. Electoral Systems: A Comparative Introduction. 2nd ed. New York: Palgrave Macmillan. Ferrara, Federico. 2004. "Electoral Coordination and the Strategic Desertion of Strong Parties in Compensatory Mixed Systems with Negative Vote Transfers." Electoral Studies 23 (3): 391-413. Ferrara, Federico, Erik S. Herron, and Misa Nishikawa. 2005. Mixed Electoral Sys- tems: Contamination and Its Consequences. New York: Palgrave Macmillan. Flanagan, Scott C., Shinsaku Kohei, Ichiro Miyake, Bradley M. Richardson, and Joji Watanuki. 1991. The Japanese Voter New Haven: Yale University Press. Fujimura, Naofumi. 2007. "The Power Relationship between the Prime Minister and Ruling Party Legislators: The Postal Service Privatization Act of 2005 in Japan." Japanese Journal of Political Science 8 (2): 233-61. Gaunder, Alisa. 2007. Political Reform in Japan: Leadership Looming Large. London: Routledge. Giannetti, Daniela, and Bernard Grofman. 2011. "Introduction: Long-Run Con- sequences of Electoral Rules Change: Comparing Italy and Japan." In A Natu- ral Experiment on Electoral Law Reform: Evaluating the Long Run Consequences of 1990s Electoral Reform in Italy and Japan, ed. Daniela Giannetti and Bernard Grofman, 1-12 . New York: Springer. Golder, Matt. 2005. "Democratic Electoral Systems around the World, 1946- 2000." Electoral Studies 24 (1): 103-21. Greene, William H. 2012. Econometric Analysis. 7th ed. Essex, England: Pearson Education. Hayao, Kenji. 1993. The Japanese Prime Minister and Public Policy. Pittsburgh: Uni- versity of Pittsburgh Press. Herron, Erik S., and Misa Nishikawa. 2001. "Contamination Effects and the Num- ber of Parties in Mixed-Superposition Electoral Systems." Electoral Studies 20 (1): 63-86. Hicken, Allen, and Yuko Kasuya. 2003. "A Guide to the Constitutional Structures and Electoral Systems of East, South and Southeast Asia." Electoral Studies 22 (1): 121-51. Horiuchi, Yusaku, and Masaru Kohno. 2004. "SNTV and Its Political Conse- quences: Mechanical Effect, Strategic Candidate Nomination, and the Origins of Japan's One-Party Dominance." Paper delivered at the 2004 Annual Meeting of American Political Science Association, Chicago, September 2-5. Huang, Chi. 2008. "Contexts and Consequences of Electoral Systems" [in Chi- nese]. In Ruhe Pinggu Xuanzhi Bianqian: Fangfalun de Tantao [The consequences of electoral system change: Methodological perspectives], ed. Chi Huang and Chin-hsing Yu, 1-18. Taipei: Wunan Press. Huang, Chi. 2010. "Causal Inference and Treatment Effect Evaluation: Partial Identification Approach and Its Application to Electoral System Effect" [in Chinese]. Xuanju Yanjiu [Journal of Electoral Studies] 17 (2): 103-34. Huang, Chi. 2011. "Political Consequences of the MMM Electoral Systems in Tai- 48 Mixed-Member Electoral Systems in Constitutional Context wan and Japan." Paper presented at the 2011 Annual Meeting of the American Political Science Association, Seattle, September 1-4. Huang, Chi. 2013. "Taiwan's Political Geography Information System (TPGIS)" http://tpgis.nccu.edu.tw/nccu/ (accessed on March 1, 2013). Huang, Chi, Lu-huei Chen, and Ying-lung Chou. 2008. "Taiwan's New Mixed Electoral System and Its Effects on 2008 Legislative Election" Presented at the 2008 Annual Meeting of the Japanese Association of Electoral Studies, Tokyo, May 17-18. Huang, Chi, Ding-ming Wang, and Ming-feng Kuo. 2008. "Straight- and Split- Ticket Voting in a Mixed-Member Majoritarian System: An Analysis of the 1996 House Election in Japan" [in Chinese]. Xuanju Yanjiu [Journal of Electoral Studies] 15 (2): 1-36. Huang, Chi, and T. Y. Wang. 2014. "Presidential Coattails in Taiwan: An Analysis of Voter- and Candidate-Specific Data." Electoral Studies 33 (March): 175-85. Japanese National Diet Library WARP Database [in Japanese]. http://warp.ndl. go.jp/search/. Jou, Willy. 2009. "Electoral Reform and Party System Development in Japan and Taiwan." Asian Survey 49 (5): 759-85. Jou, Willy. 2010. "Toward a Two-Party System or Two Party Systems? Patterns of Competition in Japan's Single-Member Districts, 1996-2005." Party Politics 16 (3): 370-93. Kabashima, Ikuo, and Gill Steel. 2007. "How Junichiro Koizumi Seized the Lead- ership of Japan's Liberal Democratic Party." Japanese Journal of Political Science 8 (1): 95-114. Kohno, Masaru. 1997. "Voter Turnout and Strategic Ticket-Splitting under Japan's New Electoral Rules." Asia Survey 37 (5): 429-40. Kostadinova, T. 2002. "Do Mixed Electoral Systems Matter? A Cross-National Analysis of Their Effects in Eastern Europe." Electoral Studies 21 (1): 23-34. Krauss, Ellis S., and Benjamin Nyblade. 2005. "'Presidentialization' in Japan? The Prime Minister, Media, and Elections in Japan." British Journal of Political Science 35 (2): 3 57-68. Krauss, Ellis S., and Robert J. Pekkanen. 2011. The Rise and Fall of Japan's LDP: Political Party Organizations as Historical Institutions. Ithaca: Cornell University Press. Kuo, Ming-feng, Chi Huang, and Ding-ming Wang. 2012. "The Contamination Effect in a Mixed-Member Majoritarian System: The Influence of LDP's Dual Candidacy in Japanese House Elections (1996-2005)" [in Chinese]. Zhengzhi Kexue Luncong [Taiwanese Journal of Political Science] 51:161-216. Kuroda, Yasumasa. 2005. The Core of Japanese Democracy: Latent Interparty Politics. New York: Palgrave Macmillan. Laakso, Markku, and Rein Taagepera. 1979. "Effective Number of Parties: A Mea- sure with Application to West Europe." Comparative Political Studies 12 (1): 3-27. Lewis-Beck, Michael S. 1986. "Interrupted Time Series." In New Tools for Social Scientists: Advances andApplications in Research Methods, ed. William D. Berry and Michael S. Lewis-Beck, 209-40. Beverly Hills: Sage. Lijphart, Arend. 1994. Electoral Systems and Party Systems: A Study of Twenty-Seven Democracies, 1945-1990. Oxford: Oxford University Press. The Consequences of MMM on Party Systems 49 Lijphart, Arend. 1997. "The Difficult Science of Electoral Systems: A Commentary on the Critique by Alberto Penades." Electoral Studies 16 (1): 73-77. Lin, Jih-wen. 2008. "To Lose Is to Win: The Candidate-Placement Strategy of Minor Parties under Japan's Mixed-Member Majoritarian System" [in Chi- nese]. Hsuanchu Yanjiu [Journal of Electoral Studies] 15 (2): 37-66. Marsh, Lawrence C., and David R. Cormier. 2002. Spline Regression Models. Thou- sand Oaks, CA: Sage. Maeda, Ko. 2008. "Re-Examining the Contamination Effect ofJapan's Mixed Electoral System Using the Treatment-Effects Model." Electoral Studies 27 (4): 723-31. Maeda, Ko. 2010. "Factors behind the Historic Defeat of Japan's Liberal Demo- cratic Party in 2009." Asian Survey 50 (5): 888-907. Massicotte, L., and A. Blais. 1999. "Mixed Electoral Systems: A Conceptual and Empirical Survey." Electoral Studies 18 (3): 341-66. McCleary, Richard, and Richard A. Hay Jr. 1980. Applied Time Series Analysis for the Social Sciences. Beverly Hills: Sage. Moser, Robert G., and Ethan Scheiner. 2004. "Mixed Electoral Systems and Elec- toral System Effects: Controlled Comparison and Cross-National Analysis." Electoral Studies 23 (4): 575-99. Moser, Robert G., and Ethan Scheiner. 2012. Electoral Systems and Political Context: How the Effects of Rules Vary across New and Established Democracies. Cambridge: Cambridge University Press. Nishikawa, Misa, and Eric S. Herron. 2004. "Mixed Electoral Rules' Impact on Party Systems." Electoral Studies 23 (4): 753-68. Nohlen, Dieter, Forian Grotz, and Christof Hartmann, eds. 2001. Elections in Asia and Pacfic: A Data Handbook, Volumes 1 and 2. Oxford: Oxford University Press. Norris, Pippa. 2004. Electoral Engineering: Voting Rules and Political Behavior Cam- bridge: Cambridge University Press. Patterson, Dennis, and Hans Stockton. 2010. "Strategies, Institutions, and Out- comes under SNTV in Taiwan, 1992-2004."Journal ofEast Asian Studies 10 (1): 31-60. Przeworski, Adam, and Henry Teune. 1970. The Logic of Comparative Social Inquiry. New York: John Wiley and Sons. Ramseyer, J. Mark, and Frances M. Rosenbluth. 1997. Japan's Political Marketplace: With a New Preface. Cambridge: Harvard University Press. Reed, Steven R. 1994. "Democracy and the Personal Vote: A Cautionary Tale from Japan." Electoral Studies 13 (1): 17-28. Reed, Steven R. 1999. "Strategic Voting in the 1996 Japanese General Election." Comparative Political Studies 33 (2): 2 57-70. Reed, Steven R., ed. 2003. Japanese Electoral Politics: Creating a New Party System. London: RoutledgeCurzon. Reed, Steven R. 2005. "Japan: Haltingly toward a Two-Party System." In The Politics of Electoral Systems, ed. Michael Gallagher and Paul Mitchell, 277-93. Oxford: Oxford University. Reed, Steven R., and Kay Shimizu. 2009. "An Overview of Postwar Politics." In Political Change in Japan: Electoral Behavior, Party Realignment, and the Koizumi Reforms, ed. Steven R. Reed, Kenneth Mori McElwain, and Kay Shimizu, 5-2 5. Baltimore: Brookings Institution. 50 Mixed-Member Electoral Systems in Constitutional Context Reed, Steven R., and Michael E Thies. 2001. "The Causes of Electoral Reform in Japan." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg, 152-72. Oxford: Oxford University Press. Reilly, Benjamin. 2006. Democracy and Diversity: Political Engineering in the Asia- Pacfic. Oxford: Oxford University Press. Reilly, Benjamin. 2007. "Democratization and Electoral Reform in the Asia-Pacific Region." Comparative Political Studies 40 (11): 13 50-71. Richardson, Bradley M. 1988. "Constituency Candidates versus Parties in Japanese Voting Behavior." American Political Science Review 82 (3): 695-718. Rochon, Thomas. 1981. "Electoral Systems and the Basis for the Vote: The Case of Japan." In Parties, Candidates and Voters in Japan: Six Quantitative Studies, ed. J. C. Campbell. Michigan Papers in Japanese Studies, No. 2, 1-28. Ann Arbor: University of Michigan Center for Japanese Study. Sakamoto, Takayuki. 1999. "Explaining Electoral Reform: Japan versus Italy and New Zealand." Party Politics 5 (4): 419-3 8. Scheiner, Ethan. 2006. Democracy without Competition in Japan: Opposition Failure in One-Party Dominant State. Cambridge: Cambridge University Press. Scheiner, Ethan, and Filippo Tronconi. 2011. "Electoral Reform in Italy and Japan: Unanticipated Outcomes?" In A Natural Experiment on Electoral Law Reform: Evaluating the Long Run Consequences of 1990s Electoral Reform in Italy and Japan, ed. Daniela Giannetti and Bernard Grofman, 97-114. New York: Springer. Senkyoho Kenkyukai (Election Law Study Community), ed. 1995. Senkyoho to Seri shikinho tesatsu [The guide of Japanese electoral law and political funds control law]. Nagoya: Shinnippon-Hoki Publishing. Shadish, William R., Thomas S. Cook, and Donald T. Campbell. 2002. Experi- mental and Quasi-Experimental Designs for Generalized Causal Inference. Boston: Houghton Mifflin. Shiratori, Rei. 1995. "The Politics of Electoral Reform in Japan." International Political Science Review 16 (1): 79-94. Shen, Yu-chung. 2005. "Relationship between the Executive and the Legislative under the Condition of Institutions: A Case Study of Taiwan's Constitutional Framework after 1997" [in Chinese]. Zhengzhi Kexue Luncong [Taiwanese Jour- nal of Political Science] 23:27-60. Shen, Yu-chung. 2011. "The Content of Dual-Executive Systems under Semi- Presidentialism" [in Chinese]. Zhengzhi Kexue Luncong [Taiwanese Journal of Political Science] 27:33-64. Shugart, Matthew Soberg, and John M. Carey. 1992. Presidents and Assemblies: Constitutional Design and Electoral Dynamics. New York: Cambridge University Press. Shugart, Matthew Soberg, and Martin P. Wattenberg. 2001. "Mixed-Member Elec- toral Systems: A Definition and Typology." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Watten- berg, 7-24. Oxford: Oxford University Press. Stockton, Hans. 2010. "How Rules Matter: Electoral Reform in Taiwan." Social Sci- ence Quarterly 91 (1): 21-41. The Consequences of MMM on Party Systems 51 Taagepera, Rein, and Matthew Soberg Shugart. 1989. Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven: Yale University Press. Tokifumi, Mizusaki, and Mori Hiroki. 2007. Sousenkyu no tokuhyou bunseki: 1958- 2005 [Japanese general elections: 1958-2005]. Tokyo: Bokutakusha. Wang, Ding-ming, and Ming-feng Kuo. 2009. "Voting Behavior under Mixed Electoral System: A Comparison between Taiwan and Japan during the Elec- toral Transition" [in Chinese]. Xuanju Yanjiu [Journal of Electoral Studies] 16 (2): 101-30. Wang, Ding-ming, Ming-feng Kuo, and Chi Huang. 2008. "Exploration of the Duverger's Psychological Effect on Electoral Transition: The Experience of the Electoral Reform in Japan" [in Chinese]. Wenti Yu Yanjiu [Issues and Studies] 47 (3): 1-28. Wang, Yeh-li. 2011. Bijiao Xuanju Zhidu [Comparative Electoral Systems]. Taipei: Wunan. Wang, Yu. 2011. "Evaluating the Candidate Incumbency Effect in Japan's Single- Member District Elections, 1996-2005." Representation 47 (2): 171-85. Yang, Chun-chih. 2002. "Japanese Coalition Government and Political Reform in the Post-Cold War Era" [in Chinese]. Zhengzhi Kexue Luncong [Taiwanese Journal of Political Science] 16:63-88. TWO The Consequences of Constitutional Systems on Party Systems Jih-wen Lin Is the Mixed-Member Majoritarian System Creating a Two-Party System? For the competitors of a legislative election, winning a seat is only the first step toward sharing political power. How powers are distributed depends on the constitutional arrangement of executive-legislative relations, but studies on electoral systems are largely legislative-centric. An executive- centric theory of electoral systems plays an important role in filling this gap. Japan and Taiwan are perfect cases-they traveled a similar path of electoral reform but are distinguishable by constitutional systems. The fol- lowing analysis explains why these two cases are worth comparing. The first reason is the similarity of Japan and Taiwan compared with the global pattern. From a comparative perspective, the Asian model deserves special attention. As pointed out by Reilly (2007a, 2007b), the Asian expe- rience is characterized by a decrease in proportionality, while the trend in the rest of the world is just the opposite.' If an increase in proportionality results from the fragmentation of the party system, the best justification of the Asian model of electoral reform is the expectation of a more efficient and stable government. Indeed, the introduction of a majoritarian-leaning electoral system is likely to improve government stability as it will reduce the number of parties (Duverger 1964; Rae 1971; Lijphart and Grofman 52 The Consequences of Constitutional Systems on Party Systems 5 3 1984). Another effect of a majoritarian-leaning electoral system is that it facilitates power turnover when an election is competitive: when only one winner is to be elected, a party gains no seats at all if it falls one vote short of the winning party. By creating a larger seat swing, a majoritarian-leaning system makes it easier for the voters to hold the ruling party accountable. Within the Asian model, Japan and Taiwan are good cases to compare.2 Both countries are in East Asia and are embedded in a political culture that emphasizes personal connections; both traveled down a similar path of electoral reform and saw a decrease in the number of parties. The second reason is their difference in executive-legislative rela- tions. As the dependent variable, small parties play a more important role in Japan than in Taiwan, and the effective number of parties in Taiwan is decreasing faster than it is in Japan (Jou 2009). We thus have two cases of the "most similar system design" to compare. The question is how to link the party system to the constitutional system, the independent variable of this chapter. As Huang, Kuo, and Stockton detail in chapter 1, small parties-those that ranked third or lower in legislative elections-have fared far better in Japan than in Taiwan. There are some differences in the Japanese and Taiwanese mixed-member majoritarian (MMNIM) systems, but Huang, Kuo, and Stockton conclude that these variations cannot fully account for the differences in the fates of small parties in the two systems. This chapter argues that the differing constitutional systems, not the differing electoral systems, are the most important source of the contrast- ing fates of small parties. If the electoral system determines the winning threshold of each electoral district, the rule of portfolio allocation specified by the constitution affects the incentive for legislators elected in different constituencies to join a national political party. The major hypothesis of this chapter is that MMM does make it harder for weaker candidates to win in each district, but whether the winners across all districts are from two parties still depends on how the executive offices are distributed. Given Japan's parliamentary system, prime ministers are elected by the Diet. Leg- islators can thus bargain with the prime minister by threatening to defect from the ruling party-splinter parties are sometimes formed in these circumstances. In contrast, Taiwan's executive offices are allocated by the president without legislative participation, giving legislators elected in dif- ferent districts a strong motivation to stay in the national parties. Another complication is the legislative process: while Taiwan has a unicameral leg- islature, the Diet of Japan includes the House of Representatives (HR) and the House of Councillors (HC), with the latter mandated to approve the 54 Mixed-Member Electoral Systems in Constitutional Context policy bills proposed by the prime minister. It has been pointed out that bicameralism-especially using different formulas to elect the delegates of the two chambers-strengthens the leverage of small parties because inter- cameral agreement is needed to adopt a bill (Hammond and Miller 1987; Riker 1992; Brennan and Hamlin 1992; Tsebelis and Money 1997; Heller 2001; Druckman and Thies 2002). The influence of Taiwan's small parties on policy making is limited because its unicameral legislature does not give them the chance to reject government bills. How the head of state, the head of government, and the legislature interact with each other in Taiwan and Japan is presented in table 2.1. It is clear that the major difference between the two constitutional sys- tems is the rule of portfolio allocation: Taiwan's premier and ministers are appointed by the president, the de facto supreme leader, without legisla- tive participation; the Japanese prime minister is designated from among the members of the Diet by a resolution of the Diet before he appoints his cabinet ministers. Taiwan's small parties have little role to play in the unicameral legislature, whereas Japan's bicameralism gives large parties an incentive to include nonpivotal small parties as coalition partners so that the latter can help adopt the government bills in the upper house. How these constitutional systems are related to the number of parties will be discussed later. In the next section, we will show how the role of political parties in portfolio allocation is defined by the constitutional system and how Riker's size principle can be used to derive hypotheses. Sections three and four will examine how party politics interact with the constitutional systems of Taiwan and Japan, and how small parties have different capacities in the allocation of executive positions. Section five concludes by addressing the TABLE 2.1. The Constitutional System in Taiwan and Japan Taiwan Japan Head of state President: de facto leader Emperor: symbolic head Appointment of head Premier is appointed by Before appointing the cabinet of government president without legislative ministers, prime minister is consent; ministers are designated by the Diet with the appointed by president at HR playing the decisive role. premier's recommendation. Legislature Unicameralism. Legislative Yuan Bicameralism. The HR can pass can pass vote of no confidence vote of no confidence in prime in premier and veto policy minister, HC can veto most proposals. policy proposals. Source: author's analysis of the constitutions of the two countries. The Consequences of Constitutional Systems on Party Systems 5 5 general implications of this study and how the number of parties may be affected by constitutional designs. The Role of Political Parties in Portfolio Allocation in Different Constitutions Political parties coordinate both in elections and in the distribution of executive offices (Shugart 1995; Cox 1997; Hicken 2009). In elections, the electoral system is the key determinant of what a party can do. If an elec- toral system requires the candidates nominated by the same party to com- pete for the same set of voters, the party should nominate an optimal num- ber of candidates; if a party nominates only one candidate, coordination should take place before the election starts. Thus, the shift from SNTV to MNMM marks a great transformation in the electoral roles of parties. Even so, maximizing its presence in the legislature is just the first kind of coordination work a party has to undertake. Once elected, legislators have to think about how to make the best use of their influence over portfolio allocation, which is the real goal of many politicians. In democracies, port- folio allocation is an important way to make policy changes (Budge and Laver 1986). Since the number of ministerial positions to be distributed is fixed, the "minimal winning coalition" principle holds: every member in the winning coalition should make itself pivotal, so that each can maxi- mize its share of the portfolios, hence its influence over the policies to be implemented by the ministry (Riker 1962). This principle implies that the size of the coalition and the parties in it should both be minimal because unnecessary members will reduce the share each can get. The constitutional system defines what a minimal winning coalition is. The Japanese constitution makes every party a potential pivot. The high- est executive power is held by the prime minister, who is designated by both chambers of the Diet-with the HR playing the decisive role-and then formally appointed by the emperor.3 The power of Japan's small par- ties is enhanced by its bicameralism in two ways. First, it is easier for a small party to win a seat in the HC elections because the electoral system mixes 48 national PR seats with 73 SNTV seats. Second,Japan's bicameral- ism is almost symmetric because the HC can veto most bills proposed by the prime minister, which will in turn put the prime minister's leadership at stake. Small parties can thus exchange their legislative veto power for political resources. Nevertheless, Japanese prime ministers do not always want to invite small parties to join their cabinets because of a trade-off 56 Mixed-Member Electoral Systems in Constitutional Context between policy making and portfolio allocation. The prime minister can compromise with the small parties without allocating any position to their members. Whether a small party wishes to join the cabinet also depends on how much the prime minister needs its support to pass legislation. Taiwan's constitutional system arranges the allocation of ministerial positions differently. Taiwan's premier is appointed unilaterally by the popularly elected president, leaving no role for the political parties to play. According to the constitution, cabinet ministers are selected by the pre- mier, but, in practice, they are handpicked by the president. Although the legislators can launch a censure motion against the premier by rallying majority support, such a move is quite unlikely because it will put the jobs of the legislators at risk. What the president has to keep in mind is whether the legislature is controlled by an opposition majority. If this is the case, the legislators may use their veto power to prevent government bills from being adopted. Another concern is the electoral system. Under SNTV, major par- ties nominate multiple candidates in most districts, the legislators tend to deliver resources to a particular group of voters, and the parties are usually factionalized. Particularist distribution of government resources is prevalent in both MMM and SNTV, but the targets of the pork may dif- fer. It is quite common for single-member district systems to encourage pork-barrel bills targeting the undecided voters in the competitive dis- tricts rather than just a particular set of constituents. Moreover, legislators elected under MMM are supposed to represent the whole constituency, giving them the incentive to choose policy stances different from those of the central government. A shift from SNTV to M/FM thus decreases the incentives for internal factionalization. Nevertheless, there is a major difference between Taiwan and Japan. In Taiwan, the directly elected presi- dent creates more pressure for Taiwanese legislators to align their cam- paign appeals with a major national party, but Japanese legislators are freer to tailor their messages to the local district, perhaps even to the point of representing a third party. We can now hypothesize how the partisan composition of the Taiwan- ese and Japanese cabinets is affected by their respective constitutional sys- tems. First, Taiwan does not have a tradition of coalition government, and political parties have a limited influence on cabinet formation. This affects both big parties and small parties. Small parties should be shut out of the cabinet almost entirely. However, this does not mean that members of the president's party will monopolize all seats in the cabinet. The president can appoint ministers as he sees fit, and he may choose to appoint a significant The Consequences of Constitutional Systems on Party Systems 57 number of nonpartisan figures to the cabinet. Tavits (2008, 43) has argued that directly elected presidents have a preference to name nonpartisan ministers to demonstrate their status as the head of state.4 The incentive to appoint nonpartisan ministers is particularly strong when the president is leading a minority party. Rather than opting for a single-party minor- ity cabinet, which might cause gridlock, or a multiparty majority cabinet, which might weaken the president's leadership, the president may prefer to appoint nonpartisan ministers to depoliticize the cabinet and convey the image of the president as leader of the entire country. Critically, while non- partisan ministers may help the president dampen opposition from other parties in the legislature or appeal to a wider segment of the electorate, they are not representatives of or responsible to opposition parties. Alter- natively, if the primary threat to the president's power comes from factions within his own majority party, he may be more likely to emphasize his role as head of the party and prefer to appoint fewer nonpartisan ministers. The percentage of nonpartisan ministers can thus give clues to the influence of the president on cabinet formation. In sum, there is no need to test the number of parties in Taiwan's cabi- nets, because ministers are predominantly from the president's party or nonpartisans.5 Rather, nonpartisan ministers play a much more important role in highlighting the president's role in ministerial appointment. Thus, we expect that, in Taiwan, the percentage of nonpartisan ministers is higher (1) in divided government than in unified government and (2) when MMM is the electoral system than when SNTV was used. Second, political parties in Japan should exercise a greater influence on cabinet formation than their Taiwanese counterparts, for the Constitu- tion of Japan requires the prime ministers to be elected by members of the Diet. Since most cabinet ministers are concurrently legislators, we can examine how the number of parties in the cabinet can be explained by the key variables.6 The number of parties in the cabinet should increase when the cabinet is not in control of the upper house. Further, if MMM accelerates the turnover of parties in power, prime ministers should have less confidence that they will be able to maintain a majority in the HC. Thus, we expect the number of parties in the cabinet to rise after electoral reform. The logic behind these hypotheses is that the bargaining power of small parties is increased if the probability of the prime minister encountering bicameral deadlock becomes higher. The second test operationalizes the prime minister's influ- ence in cabinet formation in a slightly different way. e expect the percentage of ministers from the prime minister's party to decrease when the party of most cabinet members does not hold the upper house majority or when MMM is used, 58 Mixed-Member Electoral Systems in Constitutional Context because these factors will weaken the leadership of the prime minister for the reasons already discussed. The difference in constitutional systems permeates other political are- nas. If legislators play different roles in the allocation of executive offices, the voters' expectations of them should also diverge. Taiwan's legislative elections have become highly presidentialized and focus on the debate over national identity issues between the two largest parties. MMM reduces Taiwan's effective number of parties to less than two because the number of presidential candidates is usually two. Japan lacks a centralized national leader, and legislative candidates can compete on constituency issues, some of which may be related to the national division among the political par- ties while others are local in focus. Since the constituency cleavage plays an important role in Japan, we expect the effective number of parties to be higher than in Taiwan. What we see above are theoretical models with rich empirical implica- tions. In the next two sections, we will demonstrate how the constitutional system affects the portfolio allocation in the two cases, and therefore the number and power of small parties. We will discuss Taiwan first because the evolution of its party system shows clearly the power of the consti- tutional system despite the intervening variables that a young democracy usually faces. How Taiwan's Unicameral Semipresidential System Intermediates the Impact of Electoral Reform To validate the claim that MMM has reduced the number of parties but variations in the constitutional system give political parties different roles to play, this section will describe how Taiwan's unicameral semipresiden- tial constitution has strongly depressed the influence of political parties on cabinet formation and given the president a great deal of freedom to fill executive offices. Taiwan's constitution is parliamentary by design but highly presidential in practice.7 The seven constitutional reforms enacted since 1991 endeav- ored to make the president's de facto powers constitutional by stipulating a popularly elected president; they further enhanced the president's power by removing the Legislative Yuan's power to confirm the presidential appoint- ment of the premier. The Legislative Yuan can pass a vote of no confidence in the premier but the premier cannot initiate a snap election (Lin 2011).8 Before 2000, executive congruency gave the president the power to dismiss The Consequences of Constitutional Systems on Party Systems 59 the premier, making Taiwan a president-parliamentary regime (Shugart and Carey 1992; Shugart 2005). The president dominated the appoint- ment of the premier even when the government was divided (2000-2008). Based on this practice, no popularly elected president has consulted the legislature when appointing a premier or his cabinet ministers. By this design, the president appoints and dismisses the cabinet min- isters with the latter's accountability to the legislature in mind. The ways in which the legislators hold the cabinet accountable include a vote of no confidence-which is rarely considered-and the blocking of executive proposals-which is often applied. When explaining the president's deci- sion in portfolio allocation, MAIM is also an important factor because the majorities represented by the president in the whole nation and the legisla- tors in the MAM districts may be incongruent. To ease the pressure from MMM and opposition parties, we expect the percentage of nonpartisan minis- ters in the cabinet to follow this order: MMM plus divided government > SNTV plus divided government > MMM plus unified government > SNTV plus unified government. We do not make the number of parties in the cabinet as a vari- able because the cabinets are predominantly composed of the president's party and nonpartisans rather than the nonpresidential parties. Given the limited influence of Taiwan's small parties on cabinet for- mation, we will directly examine how the aforementioned variables affect the president's role in portfolio allocation. Table 2.2 shows how much the president dominates cabinet formation by the percentage of ministers from his party.9 Counting the change in electoral systems and the president's majority status, the data include several possible divisions, with the "MAIM plus divided government" type thus far absent. Since the percentage is a continuous variable and the independent variables are categorical, we run a one-way ANOVA by the following groupings: (1) whether a president leads a unified or a divided government, (2) electoral systems, and (3) the three presidential periods. For (3), a reasonable hypothesis is that Presi- dent Lee Teng-hui had the highest percentage of ministers from his own party because he led a unified government with legislators elected under SNTV; the lowest percentage should be found under the Chen Shui-bian presidency, as Chen never assembled a majority cabinet, especially when the Kuomintang (KMT) dominated the Legislative Yuan. Standing in between, Ma Ying-jeou's strategy represents the presidentialized unified government that MMM produces. The results in table 2.3 validate the hypotheses. Overall, the average percentage of ministers from the president's party is 58.55. If the percent- ages of ministers from the president's party are compared according to 60 Mixed-Member Electoral Systems in Constitutional Context whether the president controls the legislative majority or not, the "yes" and the "no" groups have an average of 72.3% and 44.1%, respectively. It is evident that divided government forced President Chen to appoint more nonpartisan ministers to water down the cabinet's DPP makeup so as to decrease the likelihood of his policies being rejected by the Legisla- tive Yuan. If the data are compared according to the electoral system, the "SNTV" group has 61.6% and the "MNIM" group has 49.4% on average. Finally, if the average percentages for the three presidents are compared, the "Lee period" is 96.6, the "Chen period" is 44.1, and the "Ma period" is 49.4. All one-way ANOVA tests show the expected sign, and the differences between the three presidents are not only statistically significant but also larger than the other two comparisons.'0 The cabinets headed by premiers appointed by President Ma are a salient illustration of the dominance of the president-as the head of state rather than the chairperson of the ruling party-in cabinet formation. Ma's KMT has held a clear majority of legislative seats throughout his two terms in office. However, table 2.3 shows that the percentage of ministers from the president's party, already low in 2008 compared to those of President Lee, has further decreased in the intervening years." Ma's preference for TABLE 2.2. Partisan Profile of Taiwan's Cabinet Ministers Whether largest Ministers party from controls Premier Term president's majority President Party Premier party begins party (= 1) Lee Teng-hui KMT Hau Pei-tsun KMT 6/1/1990 95.24% KMT = 1 Lee Teng-hui KMT Lien Chan KMT 2/27/1993 98.31% KMT = 1 Lee Teng-hui KMT Siew Wan-chang KMT 9/7/1997 96.30% KMT = 1 Chen Shui-bian DPP Tang Fei KMT 5/20/2000 24.32% KMT = 1 Chen Shui-bian DPP Chang Chun-hsiung DPP 10/6/2000 26.47% KMT = 1 Chen Shui-bian DPP Yu Shyi-kun DPP 2/1/2002 44.74% DPP = 0 Chen Shui-bian DPP Hsieh Chang-ting DPP 2/1/2005 62.16% DPP = 0 Chen Shui-bian DPP Su Tseng-chang DPP 1/25/2006 55.56% DPP = 0 Chen Shui-bian DPP Chang Chun-hsiung DPP 5/21/2007 51.43% DPP = 0 Ma Ying-jeou KMT Liu Chao-shiuan KMT 5/20/2008 66.67% KMT = 1 Ma Ying-jeou KMT Wu Den-yih KMT 9/10/2009 47.22% KMT = 1 Ma Ying-jeou KMT Chen Chun KMT 2/6/2012 34.21% KMT = 1 Source: Yu and Zhu (2001a, 2001b), Guoshiguan Zhengjiaochu (1994, 1998, 2001), Liu (2006), Liu (1994), and the United Daily News online database (http://udndata.com/library/), accessed March 19, 2012. Note: The number of ministers is person-time. Whether a minister comes from a particular party depends on whether he/she is (was) a member of the central committee or receives (received) paid jobs from the party, or rep- resented the party in the elections. The Consequences of Constitutional Systems on Party Systems 61 appointing nonpartisan scholars and civil servants as ministers reflects his ambivalence regarding the partisan composition of the cabinet. Given the KMT's dominance in the legislature, Ma must have minimized the role of political parties when forming his cabinets. The strong influence of the president on portfolio allocation extends to elections. Two effects can be observed. The first is that Taiwan's presiden- tial elections may have a contamination effect on legislative elections." If the two elections are held concurrently or within a short period of time, voters may use information from the highly prominent presidential race to make their decisions in the less salient legislative race. Since the presiden- tial race encourages supporters of small parties to vote strategically, they may apply the same logic to legislative elections regardless of the strength of the small party's local candidate. Presidential elections thus intensify the Duvergerian obstacles'3 facing small parties in district elections. The second effect concerns cleavage structure. Campaigning dominated by the presidential race will restrain the capacity of small parties to articulate their favorite issues. In Taiwan, small parties are forced to stand with a particular camp first of all and then distance themselves from that camp's leading presidential candidate. In the legislative election of 2012, James Soong, the leader of the People First Party, was the only presidential candidate who claimed that Taiwan should eventually be reunified with the mainland. The Taiwan Solidarity Union, at the opposite extreme, was quite insistent that Taiwan's independent sovereignty should not be infringed upon. So the two small parties could be clearly identified by their positions on national identity. That helped these small parties to consolidate their PR seats, but the overall effect was limited. What MMM does is to sharply decrease the influence of the small parties when the campaign issues are set by the large parties. TABLE 2.3. One-Way ANOVA for Partisan Compositions of Taiwan's Cabinets Mean Variance Sig. Compared by presidential No 0.441 0.024 0.052 majority Yes 0.723 0.078 05 Compared by SNTV or SNTV 0.616 0.084 0.516 MMM MMM 0.494 0.027 Compared by presidents Lee 0.966 0.000 Chen 0.441 0.024 0.002 Ma 0.494 0.027 Source: Author's calculation. 62 Mixed-Member Electoral Systems in Constitutional Context How Japan's Bicameral Parliamentary System Intermediates the Impact of Electoral Reform Taiwan can be seen as a baseline when we turn our attention to Japan, a country with a longer experience of democracy under a stable constitu- tional system. When compared with Taiwan, Japan's cabinet formation is dominated by legislative parties, giving electoral systems a more important role to play. We expect to see that the number ofparties in the cabinet increases and the percentage of ministers coming from the prime minister's party decreases after MMM is introduced, especially if the cabinet does not hold the majority of seats in the HC. A review of Japan's Constitution will be helpful here.Japan's prime min- ister shall be designated by the Diet before he appoints the cabinet min- isters. According to Article 67 of the Constitution, disagreement between the two chambers about the designation of the prime minister is eventually determined by the HR. According to Article 69, the HR can pass a resolu- tion of nonconfidence in the prime minister or reject a confidence motion proposed by the latter-after which the prime minister must resign-but the prime minister can also actively dissolve the HR and call for an early election. No such relationship exists between the prime minister and the HC, but the latter can veto bills passed by the HR-except prime minister designation, treaty, and budgetary bills-unless the government can pass them a second time by a two-thirds majority in the HR. In sum, Japan's bicameral parliamentary system strengthens the bar- gaining power of the small parties when the prime minister's party is unable to control the HC. Small parties can win seats in the HC and boycott gov- ernment bills even if they are weak in the HR. Since MMM tends to create larger seat swings in the elections of the HR, the prime minister becomes less confident that he will control the majorities of both chambers. Never- theless, it is optional whether the prime minister will expand the coalition to control the majority of the HC, for reasons already explained in the theoretical section. We use Japan's portfolio allocation to test the hypotheses concerning the role of the small parties. Table 2.4 displays the partisan composition of Japan's cabinets, including the percentage of ministers from the prime minister's party and the number of parties (nonpartisan included) in the cabinets. When determining the size of the cabinet, the denominator is the number of ministers rather than their positions because a minister may concurrently hold more than one position. Note that we do not count con- trol of the HC before 1955 because that period saw a tumultuous realign- The Consequences of Constitutional Systems on Party Systems 63 ment of Japan's party system, making it difficult to measure the partisan composition of the cabinets. We should also focus on cabinets formed by prime ministers heading the largest party, otherwise the coalition govern- ments will by nature be oversized. We will use the screened data to display the role of the small parties. Since the dependent variable is continuous, we run two parts of a one-way ANOVA to test if the means are different as a result of the categories we choose. The first hypothesis is to see whether the number of parties in the Japanese cabinet varies by whether (1) the cabinet parties hold the major- ity of seats of the HC and (2) the electoral system for the HR is MAM. For the second part of the hypotheses, we will use the same categories to test how much the average percentage of ministers coming from the prime minister's party makes a difference. This is the opposite dimension of a similar variable. The empirical data are shown in tables 2.5a and 2.5b. The results largely confirm our expectation. What we see from the first half of table 2.5a is that the number of small parties in the cabinet goes up when the cabinet fails to have continuous control of the upper house. The implied message is interesting: the more the cabinet is unable to control the upper house, the greater the bargaining power of the small parties because they can threaten the prime minister and boycott his policy bills. By the same logic, the small parties lose their pivotal positions if the ruling party can unilaterally control the two chambers. The second half of table 2.5a deliv- ers a similar message in an even clearer way. Under SNTV, the prime min- isters would leave a countable number of positions for the nonpartisans rather than for the small parties because the latter are dispensable.'4 The adoption of MAM changed the strategy of portfolio allocation because the fear of losing power in an MAM race made the collaboration with some small parties inevitable, even if the number of the nonpartisan ministers remains the same. The underlying cause is that a majoritarian-leaning elec- toral system like MAM renders the undecided voters decisive and compels the prime minister to be mindful of the parties representing their interests. For the second part of the hypothesis, we use the percentage of min- isters from the prime minister's party to test the effects of the same set of variables. Note that Japan's parliamentary rule requires the prime min- isters to be elected by the Diet, with the HR playing the decisive role. The denominator is therefore "the ministers from the parties of the prime minister and the small parties in the cabinet." The average percentage of ministers from the prime minister's party when the cabinet has continuous control of the upper house is 92.9%, which is higher than the 89.1% when TABLE 2.4. Partisan Profile of Japan's Cabinet Ministers Cabinet Number of Ministers always PM'S Term non-PM from PM's controls Prime minister party begins parties party HG MA/IA/ *Shigeru Yoshida (I) JLP 5/22/1946 5 32.26% 0 0 *Tetsu Katayama JSP 5/24/1947 4 39.29% 0 0 *Hitoshi Ashita DP 3/10/1948 3 46.67% 0 0 *Shigeru Yoshida (h~a) DLP 10/15/1948 2 77.78% 0 0 *Shigeru Yoshida (Ib) LP 6/28/1950 4 44.26% 0 0 *Shigeru Yoshida (Ib) LP 10/30/1953 3 79.17% 0 0 *Shigeru Yoshida (IIb) LP 5/21/1954 1 93.33% 0 0 Ichiro Hatoyama (a) JDP 12/10/1954 1 85.00% 0 0 Ichiro Hatoyama (a) LDP 11/22/1955 1 90.91% 0 0 Ichiro Hatoyama (a) LDP 11/22/1956 1 94.74% 0 0 TanzanlIshibashi LDP 12/23/1956 1 95.00% 0 0 Nobusuke Kishi LDP 2/25/1957 1 94.59% 0 0 Nobusuke Kishi LDP 6/12/1958 1 97.73% 0 0 Hayato Ikeda LDP 7/19/1960 1 95.24% 1 0 HayatolIkeda LDP 12/8/1960 1 98.33% 1 0 HayatolIkeda LDP 12/9/1963 1 97.30% 1 0 Eisaku Sato LDP 11/9/1964 1 98.55% 1 0 Eisaku Sato LDP 2/17/1967 1 98.04% 1 0 Eisaku Sato LDP 1/14/1970 1 97.78% 1 0 Kakuei Tanaka LDP 7/7/1972 1 95.45% 1 0 Kakuei Tanaka LDP 12/22/1973 1 98.11% 1 0 Takeo Miki LDP 12/9/1974 1 92.11% 1 0 Takeo Fukuda LDP 12/24/1976 1 95.35% 1 0 Masayoshi Ohira LDP 12/7/1978 1 95.45% 0 0 Masayoshi Ohira LDP 11/9/1979 1 91.30% 0 0 Zenko Suzuki LDP 7/17/1980 1 97.56% 1 0 Yasuhiro Nakasone LDP 11/27/1982 1 91.67% 1 0 Yasuhiro Nakasone LDP 12/271983 1 98.28% 1 0 Yasuhiro Nakasone LDP 7/221986 1 95.83% 1 0 Noboru Takeshita LDP 11/6/1987 1 95.35% 1 0 Sosuke Uno LDP 6/3/1989 1 95.45% 1 0 Toshiki Kaifu LDP 8/10/1989 1 95.65% 0 0 Toshiki Kaifu LDP 2/28/1990 1 97.56% 0 0 Kiichi Miyazawa LDP 11/5/1991 1 95.65% 0 0 *Morihiro Hosokawa JNP 8/9/1993 7 18.18% 0 0 *Tsutomu Hata JRP 4/2 8/ 1994 6 39.13% 0 0 *Tomiichi Murayama SDP 6/30/1994 3 24.44% 0 0 Ryutaro Hashimoto LDP 1/11/1996 3 54.55% 1 1 Ryutaro Hashimoto LDP 11/7/1996 1 97.56% 0 1 Keizo Obuchi LDP 7/30/1998 3 84.09% 0 1 Yoshiro Moni LDP 4/5/2000 3 80.00% 0 1 Yoshiro Moni LDP 11/7/2000 3 82.35% 0 1 Junichiro Koizumi LDP 4/26/2001 4 79.49% 0 1 Junichiro Koizumi LDP 11/19/2003 2 82.86% 1 1 Junichiro Koizumi LDP 9/21/2005 2 90.91% 1 1 Shinzo Abe LDP 9/26/2006 2 88.89% 1 1 TABLE 2.4.-Continued Cabinet Number of Ministers always PM's Term non-PM from PM's controls Prime minister party begins parties party HC MNMM Yasuo Fukuda LDP 9/26/2007 2 84.38% 0 1 Taro Aso LDP 9/24/2008 3 75.00% 0 1 Yukio Hatoyama DPJ 9/16/2009 3 83.33% 0 1 Naoto Kan DPJ 6/8/2010 2 88.10% 0 1 Yoshihiko Noda DPJ 9/2/2011 2 82.22% 0 1 Source: Toshio (1990), Rei (1986, 1987), Kyofu (1996), and Naikaku seido hyakujunen kinen shi henshu iinkai (1996). In addition, the partisan composition of cabinets formed after MMM was introduced can be found at the website of the major newspapers. Notes: (1) Members of cabinet includes ministers, the chief secretary, and commission chairpersons. Nonpartisans are seen as one party when counting the non-PM parties. (2) Ministers holding concurrent positions are counted only once. (3) Whether a cabinet always controls the HC is determined by the whole term of a prime minister. For this reason, "0" may include prime ministers with interrupted majority sup- ports from the HC. (4) Prime ministers indicated by an asterisk (*) are not from the LDP or the DPJ and are excluded from analysis, except the first cabinet of Ichiro Hatoyama, who became the first LDP prime minister in November 1955. Cabinets resulting from party switching are also not counted. (5) A coalition government is defined by whether the cabinet includes at least two parties (nonpartisans excluded). TABLE 2.5A. One-Way ANOVA for the Number of Coalition Parties in Japan's Cabinets Mean Variance Sig. Compared by whether the No 1.762 0.991 0.050 coalition parties always Yes 1.250 0.303 control the upper house Compared by SNTV or SNTV 1.000 0.000 MMM MMM 2.500 0.577 0.000 Source: Author's calculation. TABLE 2.5B. One-Way ANOVA for the Ratio of Ministers from PM's Party in Japan's Cabinets Mean Variance Sig. Compared by whether the No 0.891 0.005 0.160 coalition parties always Yes 0.929 0.010 control the upper house Compared by SNTV or SNTV 0.953 0.001 MMM MMM 0.824 0.009 0.000 Source: Author's calculation. 66 Mixed-Member Electoral Systems in Constitutional Context it lacks continuous control. This result suggests an interesting dynamic: the prime minster's influence on cabinet formation declines when the cabinet fails to control the upper house. A likely cause is that bicameral deadlock weakens the prime minister's position in his cabinet. If so, electoral reform should have a significant impact because MMM is supposed to promote policy debates among political parties. Indeed, 95.3% of the ministers are from the prime minister's party when SNTV is used; this is reduced to 82.4% after MMM is introduced. In this sense, electoral reform enhances the bargaining power of the small parties on policy making even though their representatives are reduced in number. To highlight the growing influence of Japan's small parties, consider the example of the cabinet organized by Prime Minister Yukio Hatoyama (2009-10). Hatoyama's Democratic Party of Japan (DPJ) gained 308 of the 480 seats (64.17%) in the HR election of 2009 but still invited the heads of two small parties to join his cabinet: Shizuka Kamei from the People's New Party (PNP) as the minister of state for financial services and minister of state for postal reform, and Mizuho Fukushima from the Social Democratic Party (SDP) as the minister of state for consumer affairs and food safety, social affairs, and gender equality. Especially noteworthy is Fukushima, who was a member of the HC at that time. The DPJ chose to offer cabinet positions to the two smaller parties and they chose to accept them because there was a degree of policy congruence among the three parties, making it easier for them to garner support from the upper house to form a temporary majority. However, the alliance was fragile, and Fuku- shima resigned her position in 2010 over the issue of the Marine Corps Air Station Futenma.5 When the degree of policy congruence was high, the SDP was able to use its position in the upper house to win a cabinet seat; when the degree of policy congruence decreased, the SDP decided to withdraw from the cabinet and dampen the credibility of the Hatoyama cabinet. Thus this example neatly illustrates the opportunities small par- ties are afforded in Japan's constitutional system as well as the trade-offs among cabinet portfolios, policy positions, and support in the parliament that large parties much consider. Japan's party system is further fragmented by its parliamentary elec- tions. Because Japan lacks a central leader, MMM reduces the differences between political parties and gives some politicians a reason to establish their own parties. For many voters, the party platforms are intangible-at least more intangible than real politicians-and contenders in each con- stituency tend to embody their ideas through their personal image or social connections. One trend has been the formation of small parties by The Consequences of Constitutional Systems on Party Systems 67 politicians who are not very different from their fellows in the parties from which they originated. Such a trend is further endogenized by the fact that the new electoral system is conducive to regime turnover (at the national and local levels). This helps explain why some local potentates have an incentive to defect from their parties as long as they have a chance of win- ning the seat. Thus, the lack of a centralized leader makes some small par- ties powerful enough to grab the SMP seats. Nevertheless, MMM is not exactly an SMP system: the latter gives elec- toral contenders much less motivation to build their own parties, whereas the former creates contamination effects that small parties can utilize to gain some seats. Taiwan also adopted MMM, but the role of its small par- ties is much less salient than that of small parties in Japan. As we have seen, this difference cannot be fully accounted for by the design of the electoral system. It is the constitutional system that explains the varying amounts of bargaining power of the small parties. Conclusion Few topics can be studied as rigorously and fruitfully as those concern- ing electoral systems. Concepts like the threshold of winning and effective number of parties apply to all electoral systems and are well recognized in the literature. MAM, by giving priority to single-member district com- petitions, is expected to decrease the number of parties. This expectation finds support in Japan and Taiwan, both of which shifted their electoral systems from SNTV to the majoritarian-leaning MMM systems. Although the effective number of parties is smaller when measured in relative terms, the absolute number of parties in Japan is always higher than that of Tai- wan, leaving a puzzle to be explained. This chapter takes an executive-centered approach to answer this puz- zle. A theory of portfolio allocation suggests that a constitutional system instituted with more players who can veto the appointment of the head of government, and hence portfolio allocation, makes small parties more influential. Exogenously imposed, the rule of portfolio allocation gives small parties an endogenous interest in breaking away from the large par- ties in and after the legislative election. This theory explains why Japan's bicameral parliamentary system creates bargaining power for Japan's small parties, and why Taiwan's unicameral semipresidential system makes the small parties much less effective than the president when executive offices are to be arranged. Furthermore, Japan's parliamentary system gives some 68 Mixed-Member Electoral Systems in Constitutional Context electoral competitors the justification to run a campaign at the district level because the party platform is less visible than a real candidate. The chances of Japan's small parties receiving seats are thus higher than those of their counterparts in Taiwan. The partisan composition of cabinets in Taiwan and Japan corrobo- rates this argument. For Taiwan, we expect the percentage of nonpartisan ministers to be increased by MMM and the president's inability to control the Legislative Yuan, for both will increase the likelihood of the executive proposals being rejected, and nonpartisan ministers can buffer such pres- sure. The one-way ANOVA dividing the data by three Taiwanese presi- dents offers a significant count of the percentages of nonpartisan ministers. That is, Taiwan's constitutional system makes the president the paramount leader subject to the different degrees of legislative distrust of the exec- utive officers, for which electoral reform plays an important role. Small parties are almost entirely cut out of cabinet positions, as the semipresi- dential system denies them any opportunity to negotiate for ministerial posts. In Japan's parliamentary system, the HR approves the appointment of the prime minister and the HC can veto most policy bills. Since the prime minister should receive majority support from the HR, whether the cabinet also controls the HC becomes a major determinant of the cabinet's partisan composition. We expect the number of parties in the cabinet to increase and the percentage of ministers from the prime minister's party to decline when the cabinet fails to control the upper house majority. If MA/FM reduces factionalism and speeds up power transitions, we expect the electoral reform to increase the number of parties in the cabinet and decrease the percentages of ministers from the prime minister's party. These hypotheses are confirmed, and the effects of MAMM are especially impressive. Comparing the partisan compositions of the cabinets of the two coun- tries, we find Japan to have far more ministers from the prime minister's party than Taiwan has ministers from the president's party. The weights of the other parties are high in Japan while Taiwan's president tends to fill the executive positions by selecting ministers without partisanship or those from his party instead of parties with similar ideologies. This result sup- ports the theory of portfolio allocation. Taiwan and Japan are specific cases with general implications. We can make a further generalization by considering the two extreme possibilities: (1) a unicameral presidential system where the president heads the govern- ment and is the only agent responsible for portfolio allocation, and (2) a multichamber parliamentary system where the appointment of ministers The Consequences of Constitutional Systems on Party Systems 69 has to be agreed to by the political parties in all chambers. In the first case, the number of nonpartisan ministers should be high; in the latter, political parties play decisive roles in cabinet formation and the number of minis- ters from the prime minister's party depends on to what extent the prime minister controls the chambers. What electoral reform does is to alter the number of parties the political leader can choose from rather than the rules regulating portfolio allocation. That is why we need to consider the elec- toral system and the constitutional system together to explain a country's party system. NOTES 1. One should not equate the decrease in proportionality with the shift from proportional representation to a majoritarian system (see Reilly 2007a, 1359). In Reilly's cases, Cambodia switched from closed-list PR to the closed-list PR with the highest average method, Indonesia changed from closed-list PR to open-list PR, the Philippines substituted plurality-block with NMMM, and Thailand replaced block vote with MM. Only in South Korea,Japan, and Taiwan was there a change from the single nontransferable vote (SNTV) to MMM. 2. The Republic of Korea (South Korea) also abandoned SNTV (two-member district) in 1988 and replaced it with MMM, under which the voters could cast only one ballot. (The Republic of Korea shifted to a two-ballot system in 2004). Unfor- tunately for this volume, South Korea's electoral reform took place at the same time that this country transitioned to democracy, making it difficult to untangle the effects of electoral reform and democratization. For details, please see Brady and Mo (1992) and Mo and Brady (1999). In addition to the three East Asian cases, Afghanistan, Vanuatu, and Jordan have also had experience with SNTV for their national elections. 3. Sometimes the ruling party has an effective majority when it is just a few seats away from holding the majority of seats and can cooperate with the independent legislators to pass bills. If the ruling parties are unable to create an effective major- ity, the typical solution is to form a coalition government controlling the majority of seats in the HR. 4. There is also empirical evidence showing that presidential regimes are much more likely to appoint nonpartisan ministers than their parliamentary counterparts, with semipresidential regimes in between (Neto and Samuels 2010, 14). 5. The number of ministers from parties allied with the president is negligibly small. Additionally, Taiwan's ministers cannot concurrently hold a legislative posi- tion, which gives the president a strong justification to handpick the nonpartisans. 6. Coalition government is a normal practice in a parliamentary system. Although some cabinet positions are kept for nonpartisan ministers, most ministers are either from the prime minister's party or the parties of the coalition partners. 7. The president's role in the original constitution is close to a symbolic head of state. The highest administrative organ, the Executive Yuan, is headed by the premier rather than the president (Art. 53). The presidential appointment of the 70 Mixed-Member Electoral Systems in Constitutional Context premier should have the consent of the Legislative Yuan (Art. 55); a premier failing to veto a legislative resolution "shall either accede to the Legislative Yuan's view or tender his (her) resignation" (Art. 57). The president appoints the cabinet ministers on the recommendation of the premier and cannot attend the weekly meetings of the Executive Yuan. 8. These designs did not fully realize the proclaimed goals, as the high threshold of constitutional amendments required cross-partisan collaborations. 9. Due to Taiwan's authoritarian history, many people formally joined the Nationalist Party (Kuomintang, KMT) but later allowed their membership to lapse. We thus define party membership by one's active position in the party hierar- chy. Party members are those who had a seat in the central committee, held a paid job in the party bureaucracy, or represented the party in an election. 10. The president's approval rate can be another factor affecting the appointment of ministers. Presidents with high popularity are less likely to face obstruction from the legislature (Neustadt 1990; Kernell 1997). Both President Chen and President Ma saw severe declines in their approval ratings over their time in office. The per- centage of nonpartisan ministers has steadily increased under Ma as his ratings declined. Chen appointed a significant number of nonpartisan members after he was elected in 2000 to ease worries about Cross-Strait stability. As these fears eased, the number of nonpartisan members increased, but it decreased during his second term as his approval ratings dropped. Relative to his two successors, President Lee enjoyed much higher approval ratings, and the percentage of nonpartisan ministers in his cabinets was much lower. 11. Premier Jiang Yi-Huah (2013-), the successor of Chen Chun, did not renew his KMT membership until he was about to be appointed premier. Very few of Jiang's ministers are active KMT members. 12. Batto, Kim, and Matukhno (chapter 10, this volume) make a similar argument about how presidential races contaminate legislative elections. 13. For detailed examinations of Duvergerian logic, see chapter 1 and chapter 7, this volume. 14. All Japanese cabinets have nonpartisan ministers. Since we count nonparti- sans as a group, whether they are seen as a constituent part of the cabinets makes no statistical difference. 15. The U.S. base on Okinawa has been controversial because the military base is close to a populated area and creates noise pollution. The base is also a reminder that, when the Second World War came to an end, Okinawa suffered from the most deadly strike launched by the United States. REFERENCES Brady, David, and Jongryn Mo. 1992. "Electoral Systems and Institutional Choice: A Case Study of the 1988 Korean Elections." Comparative Political Studies 24 (4): 405-29. Brennan, Geoffrey, and Alan Hamlin. 1992. "Bicameralism and Majoritarian Equi- librium." Public Choice 74 (2): 169-80. Budge, Ian, and Michael Laver. 1986. "Office Seeking and Policy Pursuit in Coali- tion Theory." Legislative Studies Quarterly 11 (4): 485-506. The Consequences of Constitutional Systems on Party Systems 7 1 Cox, Gary W. 1997. Making Votes Count. New York: Cambridge University Press. Druckman, James N., and Michael E Thies. 2002. "The Importance of Concur- rence: The Impact of Bicameralism on Government Formation and Duration." American Journal of Political Science 46 (4): 760-71. Duverger, Maurice. 1964. Political Parties: Their Organization andActivity in the Mod- ern State. 3rd ed. London: Methuen. Guoshiguan Zhengjiaochu. 1994. Zhonghua minguo xingxian zhengfu zhiminglu Vol. 3 [Directory of government officials in the ROC since the implementation of the Constitution, vol. 3]. Taipei: Guoshiguan. Guoshiguan Zhengjiaochu. 1998. Zhonghua minguo xingxian zhengfu zhiminglu Vol. 4 [Directory of government officials in the ROC since the implementation of the Constitution, vol. 4]. Taipei: Guoshiguan. Guoshiguan Zhengjiaochu. 2001. Zhonghua minguo xingxian zhengfu zhiminglu Vol. 5 [Directory of government officials in the ROC since the implementation of the Constitution, vol. 5]. Taipei: Guoshiguan. Hammond, Thomas H., and Gary J. Miller. 1987. "The Core of the Constitution." American Political Science Review 81 (4): 1155-74. Heller, William B. 2001. "Political Denials: The Policy Effects of Intercameral Par- tisan Differences in Bicameral Parliamentary Systems."Journal of Law, Econom- ics, and Organization 17 (1): 34-61. Hicken, Allen. 2009. Building Party Systems in Developing Democracies. New York: Cambridge University Press. Jou, Willy. 2009. "Electoral Reform and Party System Development in Japan and Taiwan: A Comparative Study." Asian Survey 49 (5): 759-85. Kernell, Samuel. 1997. Going Public: New Strategies of Presidential Leadership. Wash- ington, DC: CQ-Roll Call Group Books. Kyofu, Utsu. 1996. Saishin rekidai naikaku soran [A compendium of the updated chronicle of cabinets]. Tokyo: Kabushiki kaisha jepi tsushinsha. Lijphart, Arend, and Bernard Grofman, eds. 1984. Choosing an Electoral System: Issues and Alternatives. New York: Praeger. Lin, Jih-wen. 2011. "The Rules of Electoral Competition and the Accountability of Semi-Presidential Governments." In Semi-Presidentialism and Democracy, ed. Robert Elgie, Sophia Moestrup, and Yu-Shan Wu. London: Palgrave Macmillan. Liu, Guoming. 2006. Zhongguo Guomindang Bainian Renwu Quanshu [A complete directory of important figures in one hundred years of the Kuomintang]. Bei- jing: Tuanjie Chubanshe. Liu, Weikai. 1994. Zhongguo Guomindang Zhiming Lu [A directory of Kuomintang party officials]. Taipei: Zhongguo guomin dang zhongyang weiyuanhui dangshi weiyuanhui. Mo, Jongryn, and David Brady. 1999. "The SNTV and the Politics of Electoral Systems in Korea." In Elections and Campaigning in Japan, Korea, and Taiwan: Toward the Study of Embedded Institutions, ed. Bernard Grofman, Sung-Chull Lee, Edwin A. Winckler, and Brian Woodall. Ann Arbor: University of Michi- gan Press. Naikaku seido hyakujunen kinen shi henshu iinkai. 1996. Naikaku seido hyakunen shi, II tsuiroku [The cabinet system in one hundred years, appendix II]. Tokyo: Okura sho insatsu kyoku. 72 Mixed-Member Electoral Systems in Constitutional Context Neto, Octavio Amorim, and David J. Samuels. 2010. "Democratic Regimes and Cabinet Politics: A Global Perspective." Revista Ibero-americana de Estudos Leg- islativos 1 (1): 10-23. Neustadt, Richard E. 1990. Presidential Power and the Modern Presidents: The Politics of Leadership from Roosevelt to Reagan. New York: Free Press. Rae, Douglas W. 1971. The Political Consequences of Electoral Laws. New Haven: Yale University Press. Rei, Shiratori. 1986. Shinpan: Nihon no naikaku III [A new edition: The cabinets of Japan III]. Tokyo: Kabushiki kaisha shin hyoron. Rei, Shiratori. 1987. Shinpan: Nihon no naikaku II [A new edition: The cabinets of Japan II]. Tokyo: Kabushiki kaisha shin hyoron. Reilly, Benjamin. 2007a. "Electoral Systems and Party Systems in East Asia." Jour- nal of EastAsian Studies 7 (2): 185-202. Reilly, Benjamin. 2007b. "Democratization and Electoral Reform in the Asia- Pacific Region." Comparative Political Studies 40 (11): 13 50-71. Riker, William H. 1962. The Theory of Political Coalitions. New Haven: Yale Univer- sity Press. Riker, William H. 1992. "The Justification of Bicameralism." International Political Science Review 13 (1): 101-16. Shugart, Matthew Soberg. 1995. "The Electoral Cycle and Institutional Sources of Divided Presidential Government." American Political Science Review 89 (2): 327-43. Shugart, Matthew Soberg. 2005. "Semi-Presidential Systems: Dual Executive and Mixed Authority Patterns." French Politics 3 (3): 323-51. Shugart, Matthew Soberg, and John M. Carey. 1992. Presidents and Assemblies: Constitutional Design and Electoral Dynamics. New York: Cambridge University Press. Tavits, Margit. 2008. Presidents with Prime Ministers: Do Direct Elections Matter? Oxford: Oxford University Press. Toshio, Otaka. 1990. Seiika jinmeijiten [A dictionary of politician names]. Tokyo: Nichigai asoshietsu kabushiki kaisha. Tsebelis, George, and Jeannette Money. 1997. Bicameralism. New York: Cambridge University Press. Yu, Keli, and Xianlong Zhu. 2001a. Zhongguo guomindang quanshu I [A complete directory of the Kuomintang I]. Xian: Shaanxi Renmin Chubanshe. Yu, Keli, and Xianlong Zhu. 2001b. Zhongguo guomindang quanshu II [A complete directory of the Kuomintang II]. Xian: Shaanxi Renmin Chubanshe. THREE LDP Factions under SNTV and MMM Yoshiaki Kobayashi and Hiroki Tsukiyama The Liberal Democratic Party (LDP) of Japan is notable as one of the most factionalized parties in modern democracies. For a long time, many scholars have discussed why LDP legislative factions (habatsu) appeared and why they are highly institutionalized. The literature offers two different explanations for the basis of LDP factions (Kohno 1992; Cox, Rosenbluth, and Thies 2000). One stresses the uniqueness of Japanese political culture including the affinity for patron-client relationships (oyabun-kobun kankei) between senior bosses and young followers (Thayer 1969; Richardson and Flanagan 1984; Baerwald 1986); the other stresses the rational incentives of LDP members (Sato and Matsuzaki 1986; Kohno 1992; Ramseyer and Rosenbluth 1993; Cox and Rosenbluth 1993, 1996; Cox, Rosenbluth, and Thies 2000). Over the last two decades, the rational choice school has become the dominant explanation. The rational choice explanation for LDP factionalism particularly focuses on the old electoral system of Japan's lower house, the single nontransferable vote system. Under SNTV, multiple seats are open in a single district but each voter can cast only a single vote, so parties seeking governmental power have to nominate multiple candidates in districts and candidates affiliated with the same party have to scramble for that single vote. Some scholars argue that LDP factions are nothing more than an instrument to help their members in this intraparty struggle. Specifically, factions have supplied their members with some electoral resources: endorsements, financial support, and cabinet and party posts. 73 74 Mixed-Member Electoral Systems in Constitutional Context In return, their members have sworn loyalty to the faction leaders in the party's presidential election (Sato and Matsuzaki 1986; Ramseyer and Rosenbluth 1993). In terms of electoral systems, legislative factionalism should be trans- formed by electoral reform. In the Japanese case, the SNTV system used in lower house general elections was replaced with a mixed-member majori- tarian (MMM) system in 1994. Japan's MMM is the combination of 300 seats elected from single-member districts (SMDs) and 180 seats1 elected from a closed-list proportional representation (PR).2 Under MAM, which centers on a single member plurality system, parties nominate a single can- didate in each district and they are no longer threatened by friendly fire from within the same party, so the electoral presence of factions should fall off. Hence, LDP factionalism should theoretically be reduced due to the electoral change from SNTV to MMN in 1994. Unlike the theoretical expectation, however, old factionalism in the LDP still exists after the electoral reform. Why didn't LDP factions imme- diately fade away? The recent literature suggests that the reason for the continuing existence of LDP factions is that they remain influential in allo- cating posts, even if they have lost power in selecting candidates and the LDP presidency (Cox, Rosenbluth, and Thies 1999; Park 2001; Reed and Thies 2001b; Krauss and Pekkanen 2004, 2010). We roughly follow this argument, but the previous studies have overlooked one side of the close relationship between the decay (and persistence) of factions and the leader- ship of the party headquarters. Our findings suggest that the reasons for and consequences of factional decline and persistence must be understood in the context of the party president's leadership within Japan's parliamen- tary system. In this chapter, we examine the long-term transition of LDP factions in the lower house under SNTV and MMM. Specifically, we investigate the change of two functions of factions from 1973 to 2009: (1) recruit- ment of new Diet members, and (2) allocation of cabinet posts. Our analyses show that the electoral and personnel importance of factions have weakened, and that these roles have been centralized in the LDP president; but that the power of the president has been unstable depend- ing on fluctuations in his popularity. Finally, we discuss the fluidity of the president's leadership in terms of the constitutional system in Japan, a parliamentary system in which a prime minister always has to maintain a majority in the Diet. LDP Factions under SNTV and MMM 7 5 Previous Studies LDP Factions under SNTV We begin by reviewing the previous studies of LDP factions under SNTV. The prevailing view of the causes of LDP factionalism is based on the ratio- nal choice model, which attributes the institutionalization of factions to the rational choices of their members. In particular, the electoral incentives of both LDP candidates and seekers of the LDP presidency are crucial. Under the old SNTV electoral system, the LDP nominated more than one candidate in districts. However, their candidates were not supported by the LDP itself on a large scale. Hence, factions, rather than party headquarters, played a substantial role in national elections. On the other hand, presiden- tial aspirants had powerful incentives to institutionalize their own factions, because the LDP president is elected by votes of LDP Diet members and local chapters (Sato and Matsuzaki 1986, 53). Factions supplied their members with some resources under the old electoral system. First, factions had been the influential actor for the endorsement of LDP candidates. "For politician-hopefuls seeking LDP endorsement, factional backing is virtually indispensable. Faction members get first priority when the party leadership hammers out its endorsement list" (Ramseyer and Rosenbluth 1993, 69). Indeed, each faction competed for the endorsement of its own faction members during the long rule of the LDP. So-called mainstream factions3 managed the LDP endorsement policy for their own candidates (Cox and Rosenbluth 1996). Second, factions furnished their members with financial support. Jap- anese politics is often related to money politics, and the important part of LDP fund-raising rested with factions. Whereas faction leaders once directly distributed their money to their followers, at the end of the 1955 System the main function of factions changed to providing "credit" and opening financial networks to their members such as fund-raising par- ties (Sato and Matsuzaki 1986, 61). Cox and Rosenbluth (1993) argue that interfactional differences in electoral fate in the 1960 and 1970s stemmed from differences in fund-raising success, but the electoral differences dis- appeared in the 1980s due to the decentralization of factional fund-raising. Third, LDP factions won cabinet and party posts from the LDP presi- dent and allocated them to their members. It is known that cabinet and party posts were distributed among factions roughly in proportion to their number in the Diet.4 The factional proportionality system for person- nel management (habatsu kinko jinji) started in the early 1970s-during 76 Mixed-Member Electoral Systems in Constitutional Context the third Sato cabinet (Sato and Matsuzaki 1986)-or in the mid or late 1970s-the Miki or Ohira cabinets (Kawato 1996). When conservative and progressive strengths in the Diet became balanced in the 1970s, the LDP shifted the personnel system from mainstream overrepresentation to fac- tional proportionality in order to maintain the interfactional coalition as a minimal winning coalition (Kawato 1996). Under SNTV, LDP factions offered these three resources, which were adequate to satisfy the rational needs of members, particularly reelection (Mayhew 1974). By outsourcing election campaigns to factional organiza- tions, the LDP solved the difficult problem of dividing votes (hyo-wari). At the same time, by allocating cabinet and party posts to factions fairly, the LDP avoided critical splits in the party (Ramseyer and Rosenbluth 1993). During the SNTV era, factions persisted by performing these essential functions. LDP Factions under MMM If SNTV shaped LDP factions, the electoral reform should have reorga- nized the factional structure. Many scholars of Japanese politics argued that Japan's electoral reform in 1994 would break the importance of fac- tions. Next, we review the previous studies on the transformation of LDP factions under MA/FM. First, factions clearly lost importance in elections after the reform. Under MMM, which centers on a single-member system, parties need not engage in intraparty competition within each district nor outsource election campaigns to factions. One Diet member interviewed by Krauss and Pekkanen (2004) observed changes following the electoral reform, explaining that "ultimately things are different than in the multi- member district system" as "nobody looks after you in elections anymore" (15). In the 1996 election, factions already lost visible roles in nominating candidates in both SMDs and PR lists (Cox, Rosenbluth, and Thies 1999). Asano (2006) finds that the importance of factions has declined in the LDP endorsement process under MM and the discretion of the party head- quarters has clearly increased as a result. Second, although the fund-raising activity of factions had already been decentralized in the 1970s (Cox and Rosenbluth 1993), the LDP's financial capacity has become vulnerable following the Political Funds Control Act reform. The reform in 1994 also included political finance reform, which prohibits politicians, their fund management organization (shikin kanri dantai), and other political organizations (including factions) from receiv- ing contributions from companies. Under the new political finance law, LDP Factions under SNTV and MMM 77 only parties5 can receive contributions from companies, so the financial power is centralized to the headquarters. Reed and Thies (2001b) refer to an article in a Japanese national newspaper that said, "LDP Diet mem- bers have been heard complaining that the new system gives the party secretary-general6 too much power because he controls the flow of the party subsidies" (398). On the subject of campaign funding, some scholars report that factions have virtually stopped offering funds to their members but the secretary-general has been known to prop up feeble candidates with monetary support (Ohtake 2003, 63-67).7 Third, the voice of factions for allocating posts has weakened after the electoral reform; however, compared to its other functions, this change has come more slowly. With the declining electoral influence of factions, the party headquarters need not heed the voice of factions as they used to, so one can expect that the headquarters will allocate cabinet and party posts with less regard to the wishes of factions. We can find striking evidence in the formation of cabinets following the reform. The factional proportion- ality system for allocating cabinet posts has obviously been shaken under MMM. For example, Keizo Obuchi, the prime minister from 1998 to 2000, used his own discretion in appointing candidates to four ministerial posts in forming his first cabinet (the so-called sosai-waku). Although Obuchi allocated other ministerial posts in proportion to each factional power, he partly ignored the recommendation lists of factions (Takenaka 2006, 156). Junichiro Koizumi, the prime minister from 2001 to 2006, attempted a much more radical change. Koizumi completely ignored the recommenda- tion lists of factions, and he decided the ministerial personnel on his own in forming his first cabinet (Takenaka 2006, 156). In contrast to the radical change in ministerial personnel, however, the literature shows that factions remain influential in allocating other positions after the electoral reform. Krauss and Pekkanen (2010, chap. 5) show that the rigid seniority system and the discrimination against nonfaction members for posts, for example Diet committee executive positions and Policy Affairs Research Council executive positions, still exist under the new electoral system, although they are now weaker than before. Park (2001) explains the matter in terms of organizational management. Indeed, as he pointed out, it is hard for the party leader to manage several hundred Diet members as a single unified group. After all, the relationship between the electoral reform and the party management is indirect, so the change comes slower than other functions. Finally, factions have no longer been an indispensable key to decide the presidency. Under MMM, elections revolve around interparty competi- tion, and the fate of the party is directly connected with the talent of party 78 Mixed-Member Electoral Systems in Constitutional Context leaders. As a result, not only Diet members but also local members and supporters come to have a strong incentive to participate in the presiden- tial election (Uekami 2010). In fact, seven of eleven LDP presidential elec- tions from 1994 to 2009 after the reform have involved the popular vote from party members (toin tohyo) (Uekami 2010, 29). Recently, the popular vote has played an important role in selecting the president. It is symbolic that Koizumi, who was not a faction leader at the time, defeated Ryutaro Hashimoto, then the leader of the largest faction, due to a huge advantage in popular votes in the 2001 presidential election. Moreover, the decay of factions has reduced factional loyalty. Faction leaders now cannot always control their member votes, as can be observed, for example, in the 2007 presidential election (Krauss and Pekkanen 2010, 140-41). In sum, the recent literature concludes that LDP factions cannot select candidates and the LDP presidency as they used to due to the electoral reform from SNTV to MMIM. However, they can still influence personnel management appointments except ministerial posts, which is one of the reasons for the continuing existence of factions. Hypotheses The presence of LDP factions in elections has weakened, but they still retain their power in personnel management. While we basically follow this argument, some questions remain. The previous studies are relatively indifferent to the consequence of the decline of LDP factionalism, espe- cially the reciprocal relation between the decay (and persistence) of fac- tions and the leadership of the headquarters. If the importance of factions has declined and their power has been centralized in the party, then the party headquarters should use the power for their own benefit. Because the previous studies are aimed at the decline of factions themselves, they do not sufficiently answer the question how the decay of factions results in the ascendancy of the party headquarters, specifically the party president.8 To address this question, we examine the long-term transition of the president's success, especially in the recruitment of new Diet members and in the allocation of cabinet posts, two of the central functions of factions. We begin with the primitive hypothesis: the power of factions is central- ized with the president due to the electoral reform from SNTV to MA/M. Assuming that the president desires the rise of his own faction to enhance his power, we can rephrase the hypothesis as follows: the power of factions is centralized with the president's faction due to electoral reform. LDP Factions under SNTV and MMM 79 First, one can expect that the faction recruitment of new Diet members not only has decreased but also has been centralized to the president's fac- tion after the electoral reform. Under SNTV, each faction normally did not nominate more than one candidate in each district, so the number of factions converged to below the maximum number of seats in multimem- ber districts (Kohno 1992). This means that the faction recruitment of new Diet members would have been carried out roughly equally among the main factions before 1994.9 On the other hand, under MMM the LDP only has to nominate a single candidate in each SMD, so the president would try to find and support candidates on his own. Furthermore, because the new political finance law deprives factions of their financial resources, factions are unable to counter the discretion of the president in selecting and supporting candidates. If so, under the new electoral system, the presi- dent's faction will get priority for the recruitment of new members. Hence, we offer the first hypothesis as follows: H1: After electoral reform, the president's faction will be more suc- cessful than other factions in recruiting new members. Second, one can expect not only that the factional proportionality system for the allocation of cabinet posts has broken down but also that cabi- net posts have been allocated to favor the president's faction members. As stressed above, the electoral importance of factions should decline due to the electoral reform, and this should also cause their influence over person- nel management decisions to wane. As can be seen from the cases of Obu- chi and Koizumi, the president's discretion for the allocation of ministerial posts has obviously increased. If so, it is no wonder that the party president uses the discretion in order to benefit his own followers. Hence, we offer the second hypothesis as follows: H2: After electoral reform, the party president will appoint a dis- proportionate number of his own faction members to ministerial positions. Data We built two datasets of LDP members of the lower house: (1) the number of new members and all members of all LDP factions from 1973 to 2009; and (2) the list of cabinet posts and their faction affiliation from the same 80 Mixed-Member Electoral Systems in Constitutional Context period (our data includes 45 cabinets from second Sato to Aso).'0 With data from the 1970s, when the LDP factional structure became stable, to the 2000s, when the LDP regime ended, we can examine the changes in the LDP factional structure before and after the electoral reform in 1994. The data are gathered from various issues of Kokkai Binran (The Diet hand- book)1 (Nihon Seikei Shimbunsha 1972-2009).' Results Recruitment of New Members To test the first hypothesis, we examine the long-term transition of the fac- tion recruitment of new Diet members from 1973 to 2009. We expect that the president's faction would gain the largest number of freshmen mem- bers among all factions under MMM but not under SNTV. Figure 3.1 shows the number of new lower house members who joined the president's faction and other major factions." As we can see, the recruitment success of the president's faction under the two electoral systems differs widely. Under SNTV, the president's fac- tion was not at a distinct advantage over other major factions except for the two elections in 1979 and 1980 when Masayoshi Ohira was the president.'4 On the other hand, in all elections under MM, except the 2009 election when the LDP was defeated by the Democratic Party of Japan," the presi- dent's faction achieved the most successful recruitment by a wide margin. The president's (Obuchi) faction gained double the number of newcomers compared to the second most successful (Mori) faction in the 1996 and 2000 elections,16 and the president's (Mori) faction gained double or triple the number of newcomers compared to the second most successful (Hashimoto) faction in the 2003 and 2005 elections. Dividing the recruitment number of the president's faction by that of the other most successful faction, the aver- age of this rate is 1.22 under SNTV but it is 2.17 under MAM. We also test the first hypothesis statistically using OLS estimation. In the analysis, we estimate the effect of the factional status in the election at period t - 1 on the recruitment just before the election at period t." The results, which are displayed in table 3.1, demonstrate that the president's fac- tion under MMM has a distinct advantage for the recruitment of new Diet members even while controlling for the effects of factional size and the over- all success of the LDP. This advantage is statistically unclear under SNTV. Figure 3.2 also shows the sizes of the president's factions and other LD P 1,ctions under SNTV and MMM 8 81 (N) 25 20 4 15 10 0 f$7 ex 4 1973 1978 1982 1987 1992 1996 2001 2005 First Election after Electoral Reform * Other Major Factions - President's Faction Fig. 3.1. Faction recruitment of new lower house members major factions. As we can see, the faction controlling the party presidency clearly succeeded in developing its power under MMM but not under SNTV. Although there were some successes under SNTV for the presi- dent's faction, such as Ohira in 1979 and 1980, the power of major factions (except the smaller Miki-Kono faction) was basically balanced regardless of which faction controlled the presidency. In contrast, the president's faction was the consistent winner in developing its power under MMM. The size of the president's faction was distinctly larger than other major factions after the electoral reform. Where did the advantage of the president under MLM come from? The keys are in the two changes after the electoral reform: the party endorse- ment process and the party election strategy headquarters. Table 3.2 is reproduced from Asano (2006, 65) and shows the number and rates of LDP endorsement under SNTV and MMM." Under both systems, almost all incumbent candidates were endorsed by the LDP, so the best indicator of factional power in the endorsement process is whether the nonincumbent candidates supported by each faction were endorsed by the LDP. There is 82 Mixed-Member Electoral Systems in Constitutional Context (N) 90 80 60 50 - 40 30 20r 10 0 t.X Other Major Factions - President's Faction 1973 1978 1982 1987 1992 1996 2001 2005 First Election after Electoral Reform Fig. 3.2. Faction size: All lower house members a clear difference in the advantage enjoyed by the president's faction after reform. Under SNTV, the president's faction won 133 endorsements. This number is only 0.94 times as large as that of the secretary-general, 1.27 times that of the chairman of the general council, and 1.13 times that of the chairman of the policy research council. By contrast, under MMM, the president's faction won 62 endorsements. This number is 1.32 times that of the secretary-general, 1.82 times the chairman of the general council, and 2.58 times the chairman of the policy research council. That is to say, the advantage of the president's faction in the endorsement process has grown stronger under VLMM," causing the rise of the president's faction (Asano 2006, 90-94). In addition, the other key to understanding the centralization of LDP recruitment is the change of the election strategy headquarters.20 Table 3.3 shows the history of the election strategy headquarters members." The LDP election strategy is mainly determined by the president, secretary-general, and director-general of the election bureau (DGEB). At LDP Factions under SNTV and MMM 83 one point, the LDP had stipulated that the secretary-general could not be selected from the president's faction (so-kan bunri).22 Since the Miki cabinet, this rule had only been broken once, by the first Ohira cabinet. Junichiro Koizumi, however, evidently abolished this custom, because he appointed fellow Mori faction member Shinzo Abe as secretary-general in 2003-4. Although Koizumi also appointed Yamasaki faction leader Taku Yamasaki in 2001-3 and Yamasaki faction member Tsutomu Takebe in 2 004-6 to the secretary-general post, the Yamasaki faction was not one of the five major factions but instead one of the newer factions that had emerged since 1998. Moreover, in the Koizumi presidency, the DGEB position was allocated to the president's Mori faction, the smaller emerging Nikai faction, and the emerging Yamasaki faction. Probably, Koizumi intended to construct a president-centered election strategy headquarters. Finally, we also examine emerging factions after the electoral reform. Figure 3.3 shows the number of new lower house members who joined TABLE 3.1. OLS Estimates Predicting Faction Recruitment Dependent Variable = Faction Recruitment 1973-93 (SNTV) 1996-2009 (MMM) Coef. Robust s.e. t-value Coef. Robust s.e. t-value (CONSTANT) -5.523 3.353 -1.94 -3.605 1.123 -3.09 * LDP's Recruitment 0.126 0.078 1.91 0.092 0.020 4.51 ** Factional Size.~- 0.153 0.035 5.32 ** 0.118 0.027 4.88 ** President. 1.644 1.437 1.20 9.818 2.829 4.06 ** Secretary-General,- 0.304 1.145 0.18 0.075 0.942 0.06 Electoral Reform Adj-R2 0.387 0.811 N-obs 50 28 1973-2009 (SNTV-MMM) Coef. Robust s.e. t-value (CONSTANT) -5.020 1.113 -4.45 LDP's Recruitment 0.105 0.025 4.16 ** Factional Size. 0.151 0.028 6.97 President, 4.254 1.794 2.70 * Secretary-General,-1 0.178 0.597 0.16 Electoral Reform 0.451 0.941 0.51 Adj-R2 0.499 N-obs 78 ***p < 0.001, **p < 0.01, *p < 0.05, tp < 0.1. 84 Mixed-Member Electoral Systems in Constitutional Context emerging factions after the electoral reform. As we can see, the Yamasaki faction controlling the secretary-general (2001-3; 2004-6)/DGEB (2006) and the Nikai faction controlling the DGEB (2004-5) had a relative advan- tage for the recruitment of freshmen members among emerging factions. In particular, the Nikai faction gained a large number of new members in proportion to its size after the 2005 election. Figure 3.4 shows the sizes of TABLE 3.2. LDP Endorsement of Conservative Candidates, 1960-2005 Non- Rates of Faction Endorsement Endorsement Total Endorsement Under SNTV/MMD (1960-1990) Incumbent President's 415 4 419 99.0% Secretary-General's 443 4 447 99.1% General Council Chair's 379 3 382 99.2% Policy Research Council Chair's 358 1 359 99.2% Other (Including Nonfaction) 1166 15 1,181 99.7% Total 2,761 27 2,788 (99.8%) (average) Non-Incumbent President's 133 36 169 78.7% Secretary-General's 141 38 179 78.8% General Council Chair's 105 38 143 73.4% Policy Research Council Chair's 118 68 186 63.4% Other (Including Nonfaction) 423 943 1,366 31.0% Total 920 1,123 2,043 (45.0%) (average) Under NMMM (1996-2005) Incumbent President's 131 0 131 100% Secretary-General's 138 0 138 100% General Council Chair's 138 1 139 99.3% Policy Research Council Chair's 108 1 109 99.1% Other (Including Nonfaction) 192 50 242 79.3% Total 707 52 759 (93.1%) (average) Non-Incumbent President's 62 1 63 98.4% Secretary-General's 47 1 48 97.9% General Council Chair's 34 8 42 81% Policy Research Council Chair's 24 1 25 96% Other (Including Nonfaction) 229 255 484 47% Total 396 266 662 (59.8%) (average) Source: Asano (2006, 65). LDP Factions under SNTV and MMM 85 emerging factions. Again, the Yamasaki faction and the Nikai faction suc- ceeded in steadily increasing their power, while other emerging factions (the Kono-Aso faction and the Ozato-Tanigaki faction) relatively failed to increase their power. We can see that not only controlling the presidency but also controlling the secretary-general and DGEB would be directly connected with the fate of emerging factions under MMM. By delegat- ing power to the smaller emerging factions, Koizumi may have cleverly avoided the rise of the other five major factions. Allocation of Cabinet Posts To test the second hypothesis, we examine the allocation of cabinet posts from 1973 to 2009.23 We expect to observe not only that the faction pro- portionality system for ministers has been broken after the electoral reform but also that the presidents have come to allocate a disproportionate num- ber of ministerial positions to their own faction. Thus, in this section, we test the two questions: whether or not the faction proportionality system exists; and whether or not the president's faction is overrepresented. Figure 3.5 shows the faction disproportionality of 44 cabinets during the period in question (1973-2009).24 The value for each cabinet is measured as follows: (1) subtracting a seat share of a faction from its cabinet share; (2) adding up the absolute values of the share distance of all factions.25 In figure 3.5, the faction disproportionality has dramatically increased since 2000.26 The disproportionality of the reshuffled second Mori cabinet is 37.45, the value of the first Koizumi cabinet is 52.06, and the value of the first Koizumi cabinet (the first reshuffled) is 69.92, respectively. These val- ues are clearly higher than previously experienced, because the average of the cabinets under SNTV was only 23.74. Although it is reasonable to suggest that the high disproportionality of the Koizumi cabinet was caused by his personnel discretion (as explained above), we have to find another explanation for the rise of the dispropor- tionality in the reshuffled second Mori cabinet.27 There are two causes. The first is the central government reform in 2001. This reform, which was approved during the second Hashimoto cabinet, reduced the number of ministries and government offices from one office and 22 ministries and agencies to one office and 12 ministries and agencies. The reshuffled sec- ond Mori cabinet was the first cabinet after the adoption of this system. Because of the reduction of ministerial posts, the disproportionality of the reshuffled second Mori cabinet increased. The second cause is the rise of emerging factions. Since 1998, the intra- TABLE 3.3. LDP Election Strategy Headquarters, 1973-2009 LDP election strategy headquarters Ed. Yr. Mo. Cabinet President Sec-Gen. DGEB 46 1973 2 2nd Tanaka STT-OHT STT-OHT STT-OHT 47 1973 8 2nd Tanaka STT-OHT STT-OHT STT-OHT 48 1974 2 2nd Tanaka (1st R) STT-OHT STT-OHT STT-OHT 49 1974 8 2nd Tanaka (1st R) STT-OHT STT-OHT STT-OHT 50 1975 2 Mild MK-K KN W-KI KN W-KI 51 1976 2 Mild MK-K KN W-KI KN W-KI 52 1976 8 Mild MK-K KN W-KI KN W-KI 53 1977 2 Fukuda.T KFA-MMM IMOM-KHIK NON 54 1977 8 Fukuda.T (R) KFA-MMM IMOM-KHIK NON 55 1978 2 Fukuda.T (R) KFA-MMM IMOM-KHIK FUNADA 56 1978 8 Fukuda.T (R) KFA-MMM IMOM-KHIK FUNADA 57 1979 2 1st Ohira IMOM-KHIK IMOM-KHIK IMOM-KHIK 58 1979 8 1st Ohira IMOM-KHIK IMOM-KHIK IMOM-KHIK 59 1979 11 2nd Ohira IMOM-KHIK KN W-KI IMOM-KHIK 60 1980 2 2nd Ohira IMOM-KHIK KN W-KI IMOM-KHIK 61 1980 8 Suzuki IMOM-KHIK KN W-KI KFA-MMM 62 1981 2 Suzuki IMOM-KHIK KN W-KI KFA-MMM 63 1981 8 Suzuki IMOM-KHIK KN W-KI KFA-MMM 64 1982 2 Suzuki (R) IMOM-KHIK STT-OHT IMOM-KHIK 65 1982 8 Suzuki (R) IMOM-KHIK STT-OHT IMOM-KHIK 66 1982 12 1 st Nakasone KNW-KI STT-OHT (Omitted) 67 1983 2 1 st Nakasone KN W-KI STT-OHT STT-OHT 68 1983 8 1 st Nakasone KN W-KI STT-OHT STT-OHT 69 1984 2 2nd Nakasone KN W-KI IMOM-KHIK IMOM-KHIK 70 1984 8 2nd Nakasone KN W-KI IMOM-KHIK IMOM-KHIK 71 1985 2 2nd Nakasone (1st R) KNW-KI STT-OHT STT-OHT 72 1985 8 2nd Nakasone (1st R) KNW-KI STT-OHT STT-OHT 73 1986 2 2nd Nakasone (2nd R)KNW-KI STT-OHT KNW-KI 74 1986 8 3rd Nakasone KN W-KI STT-OHT KN W-KI 75 1987 2 3rd Nakasone KNW-KI STT-OHT KNW-KI 76 1987 8 3rd Nakasone KNW-KI STT-OHT KNW-KI 77 1987 12 Takeshita STT-OHT KFA-MMM KFA-MMM 78 1988 2 Takeshita STT-OHT KFA-MMM KFA-MMM 79 1988 8 Takeshita STT-OHT KFA-MMM KFA-MMM 80 1989 2 Takeshita (R) STT-OHT KFA-MMM KFA-MMM 81 1989 8 1 st Kaifu MK-K STT-OHT STT-OHT 82 1990 2 2nd Kaifu MK-K STT-OHT IMOM-KHIK 83 1990 8 2nd Kaifu MK-K STT-OHT IMOM-KHIK 84 1991 2 2nd Kaifu (R) MK-K STT-OHT KFA-NMM 85 1991 8 2nd Kaifu (R) MK-K STT-OHT KFA-NMM 86 1992 2 Miyazawa IMOM-KHIK STT-OHT STT-OHT TABLE 3.3.-Continued Data Source: Kokkai Binran Ed. Yr. Mo. Cabinet 87 1992 8 Miyazawa 88 1993 2 Miyazawa (R) 89 1993 8 Hosokawa 90 1994 2 Hosokawa 91 1994 8 Murayama 92 1995 2 Murayama 93 1995 8 Murayama (R) 94 1996 2 1 st Hashimoto 95 1996 8 1 st Hashimoto 96 1996 12 2nd Hashimoto 97 1997 2 2nd Hashimoto 98 1997 8 2nd Hashimoto 99 1998 2 2nd Hashimoto (R) 100 1998 8 Obuchi 101 1999 2 Obuchi (1st R) 102 1999 8 Obuchi (1st R) 103 2000 2 Obuchi (2nd R) 104 2000 8 2nd Mori 105 2001 2 2nd Mori (2nd R) 106 2001 5 1lst Koizumi 107 2001 8 1lst Koizumi 108 2002 2 1lst Koizumi 109 2002 8 1 st Koizumi (1 st R) 110 2003 2 1st Koizumi (1st R) 111 2003 8 1st Koizumi (1st R) LDP election strategy headquarters President IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK STT-OHT STT-OHT STT-OHT STT-OHT STT-OHT STT-OHT STT-OHT STT-OHT STT-OHT STT-OHT KFA-MMM KFA-MMM KFA-MMM KFA-MMM KFA-MMM KFA-MMM KFA-MMM KFA-MMM Sec.-Gen. STT-OHT STT-OHT KFA-MMM KFA-MMM KFA-MMM KFA-MMM KFA-MMM IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK KFA-MMM KFA-MMM KFA-MMM KFA-MMM STT-OHT IMOM-KHK Y Y Y Y Y Y KFA-MMM KFA-MMM KFA-MMM KFA-MMM Y Y Y Y KFA-MMM KFA-MMM KNW-KI KNW-KI KA KFA-MMM D GEB STT-OHT IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK KFA-NMIM KFA-NMIM KFA-NMIM STT-OHT STT-OHT STT-OHT KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM KFA-NMIM N N N Y KNW-KI KNW-KI IMOM-KHK IMOM-KHK IMOM-KHK IMOM-KHK Other 112 113 114 115 116 117 118 119 120 121 122 123 124 2003 2003 2004 2004 2005 2005 2005 2006 2006 2007 2007 2008 2008 2009 9 1 st Koizumi (2nd R) KFA-MMV 12 2nd Koizumi KFA-MMMV 2 2nd Koizumi KFA-MMMV 8 2nd Koizumi KFA-MMMV 2 2nd Koizumi (R) KFA-MMV 8 2nd Koizumi (R) KFA-MMV 10 3rd Koizumi KFA-MMMV 2 3rd Koizumi (R) KFA-MMV 8 Abe KFA-MM~M 2 Abe KFA-MM~M 8 Fukuda.Y KFA-MM~M 2 Fukuda.Y KFA-MM~M 8 Fukuda.Y (R) KFA-MMM 2 Aso KA Notes: Horizontal lines indicate lower house general elections. Because of the restriction of the information source, this table does not include all cabinets and all election strategy headquarters. (N10 9 8 -7 6 5 4 1 0 2000 #'* OT N 1996 2003 2007 First Election after Electoral Reform Fig. 3.3. Faction recruitment of new lower house members (emerging factions) (Note: See theo appendix for information on LDP factions' acronyms.) (-N)40 35 30 25S 20 15 10 5 0 --Y OT N 1996 2000 2003 2007 First Election after Electoral Reform Fig. 3.4. Faction size: Al members (emerging factions) LD P 1,ctions under SNTV and MMM 8 89 (%)80 60 50 40 30 20 10 After Koizumi 1 4 -@-Minister 40 0 1972 1978 1985 1990 1996 2000 First Election after Electoral Reform 2004 2008 Fig. 3.5. Faction disproportionahty for minister factional strife within LDP factions has become obvious, and this strife led to the secession of emerging factions in the LDP. For example, the Yamasaki faction seceded from the Watanabe faction in November 1998, the Kono group seceded from the Miyazawa faction in December 1999, and the Horiuchi faction seceded from the Kato faction in January 2001.'' With the reduction of allocable cabinet posts, the increase of the num- ber of factions complicates the observance of the faction proportionality system. Because the solidarity of factions has been weakened by LMM, it is reasonable to understand the new factions as indirect effects of the electoral reform. Next, figure 3.6 shows the cabinet shares of the president's faction from 1973 to 2009. If the faction proportionality system exists, then the cabi- net share of the president's faction should correspond to its seat share in the Diet. As we can see, the cabinet share of the president's faction has 90 Mixed-Member Electoral Systems in Constitutional Context (%) 35 30 20 15 10 J After Koizumi r ,fi" h C L I M ( rr M p i -p-Minister -c-LDP Diet Member 0 1 1 1972 1978 1985 1990 1996 2000 2004 2008 First Election after Electoral Reform Fig. 3.6. Cabinet shares of president's faction been disconnected from that faction's seat share following electoral reform, especially after Koizumi. Under SNTV, the seat share of the president's faction has relatively corresponded to its cabinet share. On the other hand, under MMM presidents have more often overrepresented their own fac- tion members in the cabinet.29 However, it is contrary to our theoretical expectation that the presi- dents under MIVIIVI have not always disproportionally allocated minis- terial posts to their own faction members. Some presidents failed not only to reap a surplus profit but even to secure a proportional number of cabinet posts for their faction. If the president's discretion for per- sonnel management has been strengthened by electoral reform, why doesn't the president always form the best cabinet for him? To answer this new question, we must revise the second hypothesis and reanalyze our data. LDP Factions under SNTV and MMM 91 Additional Analyses Hypothesis Our findings in the preceding section suggest that the faction proportion- ality system for ministers has been broken and the president's discretion for the allocation of cabinet posts has strengthened, but the president has not always executed the overrepresentation strategy for his own faction. This runs contrary to our original expectation. Here, we introduce a new variable to our research: the popularity of the president (prime minister). Recently, some scholars have highlighted the importance of the popularity of the prime minister under MMM (Ito 2006; Takenaka 2006). Ito (2006, 28) argues that under MMM the prime minister becomes the party's public face, and Diet members understand that a high cabinet approval rating is a public good shared by all members of the party. When Koizumi formed his cabinet, most members went along with him and even faction leaders refrained from lodging direct criticism against him.30 We agree with this sentiment entirely, but the argument has not been examined statistically. Therefore, according to the explanation, we revise the second hypothesis as follows:" H2*: After electoral reform, the party president with high popularity will appoint a disproportionate number of his own faction mem- bers to ministerial positions. We employ two control variables to predict the overrepresentation of the president's faction. Kawato (1996, 126) argues that the LDP is more likely to employ a proportional system when it is a minimum winning coalition than when it is an oversized coalition because the withdrawal of any fac- tion from the party could cause the LDP to lose its majority in the former situation but not in the latter. Thus, we employ two variables: (1) the size of the LDP in the lower house, and (2) the effective number of factions.2 Data Here, we added new data for cabinet approval rates from 1973 to 2009. The information source is various issues ofJ7ji Seron Chosa Tokuho [Jiji pub- lic opinion poll special bulletin] (Jiji Tsushinsha 1991-2009) and its compi- lations (Jiji Tsushinsha 1981, 1992). 92 Mixed-Member Electoral Systems in Constitutional Context (Z) 80 70 60- 40 30 20 After Koizumi ( I -*- Share Rate --Approval Rate 1972 1978 1985 1990 1996 2000 First Election after Electoral Reform 2004 2008 Fig. 3.7. Cabinet shares and cabinet approval rates Results Figure 3.7 shows the transition of the president's overrepresentation rates and the cabinet approval rates in the month when a cabinet is formed. The overrepresentation rate is measured by subtracting the seat share of the president's faction from its cabinet share. Both variables are standardized in figure 3.7. We can confirm the rough correspondence between both variables under VLMM, but the relationship is unclear under SNTV For example, in the first Koizumi cabinet and the first reshuffled first Koizumi cabinet, Koizumi, bolstered by high popularity, formed a bold cabinet that clearly advantaged his Mori faction members. In the second reshuffled first Koizumi cabinet, however, he came to restrain the overrepresentation of his own followers. The change from the first Abe cabinet to the reshuffled first Abe cabinet is a similar case. We test the revised second hypothesis statistically using OLS estima- tion.34 We estimate the effect of the cabinet approval rate on the cabinet LDP Factions under SNTV and MMM 93 share of the president's faction controlling its seat share.35 The results are shown in table 3.4. The results demonstrate that the prime minister with the high (low) cabinet approval rate allocates a large (small) number of ministerial positions to its own faction members under MMM, but this mechanism is not statistically significant under SNTV. Moreover, the seat share of the president's faction has a weaker relation with its cabinet share under AMIM than under SNTV. In general, the electoral reform has shifted the resource of the president's power from his power base within the party to his popularity among the nation. The findings in figure 3.7 are also examined statistically. In addition, we also investigate the causes of the faction dispropor- tionality for ministers statistically using OLS estimation.36 The results in the analysis are displayed in table 3.5. The results demonstrate that the prime minister with a large number of factions and high popularity is likely to deviate from the faction proportionality system under MAIM but not under SNTV. TABLE 3.4. OLS Estimates Predicting Cabinet Share Dependent Variable = Disproportionality for Ministert 1972-1993 (SNTV) 1996-2009 (MMM) Coef. Robust s.e. t-value Coef. Robust s.e. t-value (CONSTANT) -13.734 26.767 -0.51 38.315 13.141 2.92 * LDP's Size (%)t0.065 0.280 0.23 -0.400 0.243 -1.64 Number of Factions 1.683 2.357 0.71 -2.633 2.286 -1.15 Seat Share (%)t0.891 0.200 4.46 0.449 0.226 1.99 t Approval Rates ()t 0.056 0.105 0.53 0.300 0.057 5.22 *** Electoral Reform Adj-R2 0.340 0.490 N-obs 24 20 1972-2009 (SNTV-MMM) Coef. Robust s.e. t-value (CONSTANT) 22.582 11.529 1.96 t LDP's Size (%)t-0.312 0.186 -1.67 Number of Factions -1.324 1.419 -0.93 Seat Share (%)t0.755 0.172 4.39 Approval Rates (%) 0.179 0.057 3.12 ** Electoral Reform 1.567 1.844 0.85 Adj-R2 0.358 N-obs 44 ***p < 0.001, **p < 0.01, *p < 0.05, tp < 0.1. 94 Mixed-Member Electoral Systems in Constitutional Context To summarize, the faction proportionality system has not been main- tained under MIM. The president's discretion for the allocation of cabi- net posts has clearly strengthened, especially after Koizumi. The personnel power of the president under MMM, however, depends on his national popularity at that time. Ultimately, this effect stems from Japan's constitutional system. Although the importance of factions has declined under MMM, they are still latent actors. This is because prime ministers always have to control a majority in the Diet and factions can extract some resources from unpopu- lar prime ministers by threatening to hold a vote of no confidence or to leave the party outright. Conclusion In this paper, we examined the transformation of LDP factions under SNTV and MMN. Under MMM, in which interparty competition is the focus, the presence of factions declines because factions are primarily ori- TABLE 3.5. OLS Estimates Predicting Disproportionality Dependent Variable = Cabinet Share of the President's Faction, 1972-1993 (SNTV) 1996-2009 (MMM) Coef. Robust s.e.m t-value Coef. Robust s.e. t-value (CONSTANT) -52.055 50.639 -1.03 -86.273 21.759 -3.96 ** LDP's Size (%) 0.618 0.523 1.18 0.576 0.491 1.17 Number of Factions 8.147 4.771 1.71 11.714 2.657 4.41 *** Approval Rates ~) -0.162 0.192 -0.85 0.398 0.165 2.41 * Electoral Reform Adj-R2 0.054 0.485 N-obs 24 20 1972-2009 (SNTV-MMM) Coef. Robust s.e. t-value (CONSTANT) -87.013 19.683 -4.420 *** LDP's Size (%) 0.684 0.348 1.960 t Number of Factions 11.419 2.029 5.630 *** Approval Rates r) 0.198 0.137 1.450 Electoral Reform 4.693 3.344 1.400 Adj-R2 0.451 N-obs 44 ***p < 0.001, **p < 0.01, *p < 0.05, tp < 0.1, (1) Newey-West standard errors with one lag. LDP Factions under SNTV and MMM 95 ented toward intraparty competition. The diminution of their electoral importance also weakens their voice in personnel management. More- over, the decay of factions leads to the relative ascendancy of the president. Indeed, we confirmed that two of a faction's main functions-recruiting new members and allocating cabinet posts-have not only declined but also been centralized to the president (or the president's faction) after the electoral reform in 1994. Under MMIM, by controlling the party endorsement policy and the election strategy headquarters, the president has succeeded in recruiting freshmen members and expanding their power within the LDP. The fall of the electoral presence of factions also has indirectly caused the weakening of factions' influence on personnel management. Although once the LDP had adopted the rigid faction proportionality system, the system has bro- ken down following the electoral reform, especially after Koizumi. While the president's discretion over personnel management has increased, he cannot always form cabinets at will. Under MMM, there is a close relationship between the overrepresentation of the president's fac- tion in the cabinet and the national popularity of the president. This can be understood in terms of the interaction between MA/M and Japan's parlia- mentary system. Under a parliamentary system, the prime minister always has to control a majority in the Diet. On the other hand, under MMM, elections focus on interparty competition, so the popularity of the party president becomes directly connected with the fate of the party. While a prime minister who enjoys high popularity enjoys considerable discretion and can largely ignore factions, factions can draw concessions from less popular prime ministers by threatening a vote of no confidence or separa- tion from the party. MMM has significantly reduced the importance of factions, but they continue to exist because they remain latent actors, espe- cially for personnel management, under the parliamentary system. 96 Mixed-Member Electoral Systems in Constitutional Context APPENDIX: LIST OF LDP FACTIONS' ACRONYMS SIT-OUT: S ato-Tanaka-Takeshita-Obuchi-Hashimoto-Tsu shima KFA-MMM: Kishi-Fukuda-Abe-Mitsuzuka-Mori-Machimura IMOM-KHK: Ikeda- Maeo-Ohira- M yazawa-Kato- Horiuchi-Koga KA: Kono-Aso 01: (Kato-) Ozato-Tanigaki KNW-KI: Kono-Nakasone-Watanabe-(Eto-Kamei)-Kamei-lbuki Y: Yamasaki MK-K: Miki-Komoto-Komura * N: [Predecessor: New Conservative Party] Nikai * Five Major Factions * Emerging Factions LDP Factions under SNTV and MMM 97 NOTES 1. Prior to the 2000 election, there were 200 PR seats. 2. See Reed and Thies (2001a) on the history of the electoral reform in 1994. 3. This term simply indicates "factions which had supported the winning candi- date in the most recent party presidential contest" (Cox and Rosenbluth 1996, 263). 4. More strictly, Sato and Matsuzaki (1986, 63-66) distinguish three types of LDP personnel systems: (1) faction power proportionality method, (2) factional representation method, and (3) all members participation method. Krauss and Pekkanen (2010, 113-14) explain the three systems in English (the English equiva- lents also follow them). According to Sato and Matsuzaki, for example, the first method is applied to ministers, the second method to deputy secretary-generals, and the third method to parliamentary vice-ministers. 5. Including local chapters of parties and political fund organization (set/i shikin dantai), of which each party can establish only one. 6. The secretary-general (kanji-cho) is the second highest party post in the LDP. It administers (1) the raising and distribution of party funds, (2) the allocation of party posts and advising the president on the allocation cabinet posts, and (3) deci- sions on election strategies on the whole (Ohtake 2003, 66). 7. The financial distribution within factions, however, has not completely dis- appeared after the electoral reform. For example, the Yomiuri Shimbun reported that the Obuchi faction and Eto-Kamei faction distributed three million yen per capita to their members for immediate use in campaigning before the 2000 election (Yomiuri Shimbun, May 6, 2000, 4; cited in Asano [2006, 43]). 8. Our research on faction recruitment, however, was greatly influenced by Asano's (2006) pioneer work on the changes in the LDP endorsement process fol- lowing the electoral reform. 9. In reality, four large factions and one small faction shared candidates of each district depending on their size, because only about a third of all districts under SNTV were maximum (five) member districts (Kohno 1992, 390). 10. Five cabinets from the Hosokawa to the 1st Hashimoto in 1993-6, however, are excluded from our analysis, because four cabinets were not dominated by the LDP and one cabinet was the special case of the LDP regime during the party realignment period. The first Mori cabinet (2000) is also excluded, because this cabinet is the provisional cabinet that took over all ministers and parliamentary vice-ministers from the reshuffled Obuchi cabinet (the so-called inuki naikaku). 11. Kokkai Binran (The Diet handbook) gives information about the factional affiliation of all LDP members of the Diet. 12. We secondarily referred to Hata's work (Hata 2001). 13. There are five major factions in the LDP. They come from the factions of five leaders: Eisaku Sato, Nobusuke Kishi, Hayato Ikeda, Ichiro Kono, and Takeo Miki. All leaders except Kono became the prime minister. See the appendix to this chapter for information on the series of five major factions. 14. It is known that the first and second Ohira cabinets were faced with fierce conflict between mainstream and nonmainstream factions. Ohira's election strategy for the expansion of mainstream factions was strongly criticized by nonmainstream factions (for example, Yomiuri Shimbun, October 4, 1979, 2; October 14, 1979, 2). In 98 Mixed-Member Electoral Systems in Constitutional Context the 1980 election (Ohira died suddenly just before the poll), the relative advantage of the Ohira faction may have been caused by the sudden dissolution of the Diet in which most factions were not ready to find fresher candidates (for example, Yomiuri Shimbun, June 3, 1980, 3). 15. In the 2009 election, the number of LDP newcomers-elect was only five. 16. Although the president Yoshiro Mori (the Mori faction) contested the 2000 election, we coded the Obuchi faction as the president's faction only in the 2000 election, because the first Mori cabinet (April-July 2000) was substantially the pro- visional government from the sudden death of Obuchi. Of course, in the same way, we can understand this election as the special case that the former president's fac- tion (which controlled the secretary-general at that time) wielded the power over recruitment against the "provisional" president. 17. The dependent variable is Faction Recruitment (the number of new lower house members of the faction,) just before the t election. There are five indepen- dent variables: (1) LDP's Recruitment (the total number of new lower house mem- bers in the whole LDP at period t); (2) Factional Size (the number of all lower house members of faction. at period t - 1); (3) President (a dummy variable that is equal to 1 if faction. controls the presidency at period t - 1); (4) Secretary-General (a dummy variable that is equal to 1 if faction. controls the secretary-general at period t - 1); and (5) Electoral Reform (a dummy variable that is equal to 1 if the faction contests the t - 1 election after 1994). Our observations are 78 factions (including the same individuals) that existed from period t - i through period t. We tested the three methods: pooled OLS model; fixed effects model; and random effects model, using F test, Breusch-Pagan Lagrangian multiplier test, and Hausman test. The results of the three tests sup- ported a pooled OLS model in all equations. Robust standard errors in table 3.1 are adjusted for clustering at the faction level. 18. We have to be careful about the difference between our and Asano's data coding of the 2000 election. As stated above, we coded the factional status in the 2000 election based on the data of the reshuffled Obuchi cabinet, but Asano coded it based on the data of the first Mori cabinet. However, we can basically obtain the same results, even if Asano's data is modified according to our data coding. 19. In the 2005 election the ascendancy of the president's faction waned because Koizumi endorsed a large number of nonfaction candidates as so-called assassin candidates (shikaku koho) (Asano 2006, 88-90). Although our research with this chapter has had to omit this effect, it should be understood as an example of party centralization. 20. The LDP's organization of its election strategy is rather complicated. We focus on three positions controlling the election strategy: the president, secretary- general, and director-general of the election bureau (senkyo taisaku kyokucho, or soumu kyokucho). 21. See the appendix for information on the LDP factions' acronyms. 22. The system started with the Shiina decision (Shiina saitei) in 1974, which decided Tanaka's successor (then president Kakuei Tanaka lost his presidency because of a scandal). The party headquarters, including Etsusaburo Shiina, the vice-president of the Tanaka cabinet (1972-74), decided that Takeo Miki be appointed as the next president. At the same time, they also "reached a consensus LDP Factions under SNTV and MMM 99 on five points, including the principle that the secretary general, finance commit- tee chairman, and accounting bureau chief would not be selected from among the party president's faction" (Masumi 1995,161). 23. In this chapter, we observe only ministerial positions. Although the allocation pattern of parliamentary vice-minister, senior vice-minister, and vice-minister posi- tions is also an interesting theme, these positions have less importance in elections than ministerial positions. For example, it is known that the LDP applies an "all members participation" method (the position is allocated to all LDP members who achieve two or three reelections) to the parliamentary vice-minister (Sato and Mat- suzaki 1986, 63-66). And, here, we do not consider the career pattern and relative importance of each ministerial post (Kobayashi 2012; Ono 2012). 24. We investigated 44 cabinets from the second Tanaka cabinet to the Aso cabi- net (except the Hosokawa, Hata, Murayama, reshuffled Murayama, first Hashi- moto, and first Mori cabinets). Strictly speaking, the second Tanaka cabinet started on December 22, 1972, so some figures in this section start from "1972." 25. Both include only faction members in the lower house. 26. The upturn in the reshuffled Miyazawa cabinet (1992) was caused by the secession of the Hata faction from the Takeshita faction. Because the Hata faction failed in gaining positions in the cabinet, it extended the whole of the dispropor- tionality. 27. Indeed, at that time, the opposition parties criticized the faction proportion- ality personnel of the reshuffled second Mori cabinet (Yomiuri Shimbun. December 6, 2000, 4). 28. Actually, major members of the Kato faction accompanied the Horiuchi fac- tion, so we coded the Horiuchi faction as one of five major factions in this paper. 29. The average of the absolute distances between the cabinet share of the presi- dent's faction and its seat share in each cabinet is 3.36% under SNTV, while that is 6.07% under MMM. 30. Krauss and Pekkanen (2004) offer a similar view on Koizumi's discretion in forming a cabinet: "Moreover, as Koizumi's popularity waned, his ability to disre- gard factions in selecting even cabinet members disappeared" (16). 31. Taking a similar view, Ono (2012) offers a sophisticated analysis of the LDP portfolio allocation. He argues that LDP presidents with a weaker power base within the party or lower popularity among the nation had surrendered more cabi- net posts to other factions from 1960 to 2007, but he finds that the electoral reform in 1994 had no effect. In contrast, our analysis demonstrates that the close relation- ship between the cabinet share of the president's faction and the popularity of the president is statistically unique or more remarkable to the MMM era (at least in comparison with the 1970-90s' SNTV era when the LDP intraparty system was stable). 32. The effective number of factions is calculated by Laakso and Taagepera's for- mula (Laakso and Taagepera 1979). 33. When Koizumi and Abe formed cabinets that treated other factions more fairly, a national newspaper emphasized a stereotype of "party unity" (kyoto taisei) (Yomiuri Shimbun, September 23, 2003, 4; August 29, 2007, 4). This phrase demon- strates one side of the relationship between the prime minister's leadership and his popularity. 100 Mixed-Member Electoral Systems in Constitutional Context 34. The Breusch-Godfrey LM test indicates the absence of first-order serial cor- relation in each model. 35. The dependent variable is Cabinet Share of the President's Faction (the cabinet share of the president's faction in the lower house). There are five independent variables: (1) LDP's Size (the seat share of the LDP in the lower house); (2) Number of Factions (the effective number of LDP factions in the lower house); (3) Seat Share (the seat share of the president's faction in lower house faction members of the LDP); (4) Approval Rates (the cabinet approval rating in the month when the cabi- net is formed); (5) Electoral Reform (a dummy variable that is equal to 1 after 1994). 36. The Breusch-Godfrey LM test indicates the absence of first-order serial cor- relation in each model, except the model of the STNV era. Therefore, we report the Newey-West standard errors for first-order serial correlation in this model. The dependent variable is Disproportionality for Minister (the sum of the abso- lute distances between the cabinet share and its seat share for each faction in the lower house). The independent variables are the same as in table 3.4, except (3). REFERENCES Asano, Masahiko. 2006. Shimin Shakai ni okeru Seido Kaikaku: Senkyo Seido Kaikaku to Kohosha Rikuruto [Institutional reform in civil society: Electoral system and candidate recruitment]. Tokyo: Keio Daigaku Shuppankai. Baerwald, Hand H. 1986. Party Politics in Japan. Boston: Allen & Unwin. Cox, Gary W., and Frances M. Rosenbluth. 1993. "The Electoral Fortunes of Leg- islative Factions in Japan." American Political Science Review 87:577-89. Cox, Gary W., and Frances M. Rosenbluth. 1996. "Factional Competition for the Party Endorsement: The Case of Japan's Liberal Democratic Party." British Journal of Political Science 26:259-69. Cox, Gary W., Frances M. Rosenbluth, and Michael F. Thies. 1999. "Electoral Reform and the Fate of Factions: The Case of Japan's Liberal Democratic Party." British Journal of Political Science 29:3 3-56. Cox, Gary W., Frances M. Rosenbluth, and Michael E Thies. 2000. "Electoral Rules, Career Ambitions, and Party Structure: Comparing Factions in Japan's Upper and Lower Houses." American Journal of Political Science 44:115-22. Hata, Ikuhiko, ed. 2001. Nihon Kanryosei Sogo Jiten 1868-2000 [Japanese bureau- cracy general encyclopedia 1868-2000]. Tokyo: Tokyo Daigaku Shuppankai. Ito, Mitsutoshi. 2006. "Kantei Syudogata Seisaku Kettei to Jiminto: Koa Eguze- kuteibu no Syukenka" [The LDP and the centralization of the core executive under the Koizumi cabinet]. Revaiasan [Leviathan] 38:7-40. Jiji Tsushinsha, ed. 1981. Sengo Nihon no Seito to Naikaku: J/i Seron Chosa ni yoru Bunseki [Political parties and cabinets of postwar Japan: Analysis with Jiji public opinion poll]. Tokyo: Jiji Tsushinsha. Jiji Tsushinsha, ed. 1992. Nihon no Seito to Naikaku 1981-91: Ji Seron Chosa ni yoru Bunseki [Political parties and cabinets of Japan 1981-91: Analysis with Jiji public opinion poll]. Tokyo: Jiji Tsushinsha. Jiji Tsushinsha. 1991-2009. YJ i Seron Chosa Tokuho [Jiji public opinion poll special bulletin]. Various Issues. Tokyo: Jiji Tsushinsha. Kawato, Sadafumi. 1996. "Shinioritei Ruru to Habatsu: Jiminto ni okeru Jinji Hai- LDP Factions under SNTV and MMM 101 bun no Henka" [The seniority rule and factions: The change in post allocation of the LDP]. Revaiasan [Leviathan] special issue: 52-77. Kobayashi, Yoshiaki. 2012. Malfunctioning Democracy in Japan: Quantitative Analysis in a Civil Society. Lanham, MD: Lexington Books. Kohno, Masaru. 1992. "Rational Foundations for the Organization of the Liberal Democratic Party in Japan." World Politics 44:369-97. Krauss, Ellis S., and Robert Pekkanen. 2004. "Explaining Party Adaptation to Elec- toral Reform: The Discreet Charm of the LDP." journal of Japanese Studies 30:1-34. Krauss, Ellis S., and Robert Pekkanen. 2010. The Rise and Fall offapan's LDP: Politi- cal Party Organizations as Historical Institutions. Ithaca: Cornell University Press. Laakso, Markku, and Rein Taagepera. 1979. "Effective Number of Parties: A Mea- sure with Application to West Europe." Comparative Political Studies 12:3-27. Masumi, Junnosuke. 1995. Contemporary Politics in Japan. Trans. Lonny E. Carlile. Berkeley: University of California Press. Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale Uni- versity Press. Nihon Seikei Shimbunsha. 1972-2009. Kokkai Binran [The Diet handbook], 45- 125. Tokyo. Ohtake, Hideo. 2003. Nihongata Popyurizumu: Set/i he no Kitai to Genmetsu [Japanese populism: Hopes and disillusionments in politics]. Tokyo: Chuko Shinsho. Ono, Yoshikuni. 2012. "Portfolio Allocation as Leadership Strategy: Intraparty Bargaining in Japan." American journal of Political Science 56:553-67. Park, Cheol Hee. 2001. "Factional Dynamics in Japan's LDP since Political Reform." Asian Survey 41:428-61. Ramseyer, Mark J., and Frances M. Rosenbluth. 1993. Japan's Political Marketplace. Cambridge: Harvard University Press. Reed, Steven R., and Michael E Thies. 2001a. "The Causes of Electoral Reform in Japan." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew S. Shugart and Martin P. Wattenberg. Oxford: Oxford University Press. Reed, Steven R., and Michael E Thies. 2001b. "The Consequences of Electoral Reform in Japan." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew S. Shugart and Martin P. Wattenberg. Oxford: Oxford University Press. Richardson, Bradley M., and Scott C. Flanagan. 1984. Politics in Japan. Boston: Little, Brown. Sato, Seizaburo, and Tetsuhisa Matsuzaki. 1986.Jiminto Seiken [The LDP regime]. Tokyo: Chuo Koron Sha. Takenaka, Harukata. 2006. Shusho Shihai: Nihon Set/i no Henbo [The prime minister ruling: Japanese politics transformed]. Tokyo: Chuko Shinsho. Thayer, Nathaniel B. 1969. How the Conservatives Rule Japan. Princeton: Princeton University Press. Uekami, Takayoshi. 2010. "Senkyo Seido Kaikaku to Jiminto Sosai Sensyutsu Katei no Henyo: Ridashippu wo Umidasu Kozo to Kosei no Sokoku" [Elec- toral reform and the transformation of LDP presidential elections: Leadership, structure, and personality]. Senkyo Kenkyu [Japanese Journal of Electoral Stud- ies] 26 (1): 26-37. FOUR Executive Competition, Electoral Rules, and Faction Systems in Taiwan Nathan F Batto and Hsin-ta Huang It is sometimes assumed that politics in Taiwan mirrors politics in Japan. Taiwan is a former Japanese colony, and it inherited much of its present institutional structure from the Japanese who were largely copying their own Meiji-era institutions back home. During the latter half of the 20th century, both countries experienced a long period of single-party domi- nance, electoral reform from SNTV to MAM, the rise of a longtime opposition party to power, and fast economic growth. Further, the KMT is often assumed to be a mirror of the LDP, and both were longtime ruling parties with strong clientelistic proclivities. Thus it is not unreasonable to expect KMT factions to operate in a similar manner to the LDP factions described by Kobayashi and Tsukiyama in chapter 3 of this volume. How- ever, the KMT's internal factions have not mirrored the LDP's. Instead of national factions, the KMT factions are primarily local. Further, instead of multiple factions, there are usually two KMT factions in each area. Finally, while the LDP factions competed primarily for top party positions and parliamentary seats in SNTV elections, KMT factions are focused primar- ily on winning control of local governments. Surprisingly, it is the longtime opposition DPP's factions that look much more like the LDP's. Taiwanese factions have been shaped primarily by executive compe- tition. However, factions have not necessarily organized to capture the national executive. Rather, they have competed over the most desirable 102 Executive Competition, Electoral Rules, and Faction Systems in Taiwan 103 post that was attainable to them at the time. KMT local factions have always been oriented around control of local governments, and this has led to bifactional systems. DPP factions were initially shaped for winning control of the national party bureaucracy as well as winning seats in the legislature, and multiple factions flourished. Since the presidency became a viable goal, DPP factions have undergone significant upheavals, and most of them have been related to competition over the presidency. Electoral reform from SNTV to MMM has had a limited impact on Taiwan's factions. For one thing, KMT local factions maintained bifac- tionalism in spite of SNTV, not because of it. MMM has done little to upset the preexisting faction systems. For another, the reform covered the national legislature only and did not extend to local assemblies or internal party elections. Just as the presidency is not always the critical executive post around which factions compete, the rules used to elect the national legislature are not necessarily the only pertinent important electoral rules. In this chapter, we examine the recent evolution of KMT local fac- tions and DPP national factions. For the KMT, we focus extensively on the Taichung County Red and Black factions, which have dominated local politics since the 1950s. For the DPP, we retrace the twists and turns of the factional struggle, as various factions are born, form alliances, fight bit- terly, splinter, and disintegrate altogether. The discussion of the KMT is informed by extensive interviews with local politicians in Taichung,' while the section on the DPP relies heavily on newspaper accounts. KMT Local Factions There are multitudes of local factions in Taiwan, and it is consequently extremely difficult to make statements that apply to all of them. However, we can make a few generalizations about the characteristics and goals of almost all local factions.2 First, local factions are geographically bounded. They are formed along local administrative boundaries, and their activities are confined within this geographical area. The largest factions operate within the boundaries of a single county, while smaller factions appear in smaller administrative districts, such as townships. Second, most factions grew out of competition among local elites over political posts, such as township mayor, county assembly seats, and especially for the county mag- istrate.' This competition created two or more local factions that have been defined, in part, by their competition with and opposition to each other. Third, factions are informal structures. However, factional affiliations for 104 Mixed-Member Electoral Systems in Constitutional Context faction bosses and other prominent politicians are usually widely known. Interviews reveal very little ambiguity among insiders about who belongs to which faction. Fourth, the overriding goal that motivates factions is to secure political, economic, and social interests. This can be done by tak- ing as many political posts as possible and by cooperating with the KMT. Fifth, the factions currently compete for political posts in both the public and quasi-public sectors. The latter includes farmers, fishers, and irrigation associations, which are valuable both for their contacts with large numbers of farmers and fishers and also for the financial clout of their affiliated credit unions. However, the single most important prize is the county mag- istrate's seat. The Nature of KMT Factions There are two distinct understandings of KMT factions. One set of scholars sees factions as being constructed in a guided, top-down process through patron-client relationships. A second set of scholars understand factions as founded on a set of social relationships. In this view, factions form in a natural, bottom-up process. The first school sees factions as built on a series of unequal and hier- archical bilateral relationships cultivated from the top down. In a patron- client relationship, a patron provides instrumental benefits to a client of lower status in return for general support or personal services (Scott 1972, 92). The relationship is unequal, with a large pool of potential clients competing for a small pool of potential patrons. In the KMT factions, the KNIT sits atop four tiers of hierarchical relationships. The party acts as a patron to the faction core, who act as patrons to electoral cadres, who are patrons to faction followers. In each tier, the patron provides protection and privileges, such as nomination to particular political posts, monopolis- tic economic interests, or constituency service, to the client in exchange for political loyalty and support in elections (Wang and Huang 2010). In this view, local factions were deliberately created to satisfy the KMT's need for stable government. Local political elites, who had first come onto the scene during the Japanese era, gained increased prominence at the inception of KMT rule through local elections for posts in local gov- ernments or councils. The KMT did not interfere much in local politics during these early years, and the elites began to form horizontal and verti- cal alliances with other local politicians. As the KNIT solidified its rule in Taiwan, it began to cultivate these local elites to ensure long-term stability under the guiding principles of localizing political participation, balanc- Executive Competition, Electoral Rules, and Faction Systems in Taiwan 105 ing competing forces, and offering economic privileges (M. Chen 1995, 150; N. Wu 1987, 301-34). The new local factions were thus absorbed into the KMT's political structure, relying on the KMT for their existence and operation. The KMT resided at the top tier of this structure as the sole patron to all local factions. They could compete with each other only for the patronage of the KMT. If they went against the will of the KMT, they could be shoved aside or even be put out of existence. In this structure, the faction bosses (or core members) were usually local gentry or members of prominent families. They typically held more eco- nomic resources and had stronger motives to be politically involved. Both attributes made them desirable targets for the KMT. Faction bosses acting as patrons handled political and economic resources, distributing them to the faction's electoral cadres, their clients. Electoral cadres intermediated between local politicians and their supporters, mobilizing supporters to vote in elections in return for local political and economic resources such as low-level political posts or uncompetitive contracts for small public works projects. Electoral cadres patronized their own followings. These faction followers traded their votes for help in dealing with the govern- ment. Moreover, faction followers were generally alienated from politics due to the prevailing authoritarian political environment and did not have strong opinions about which candidate deserved their votes. At election time, they often deferred to a trusted electoral cadre for guidance, and the election cadre, as a savvy and connected observer of local politics, served as an opinion leader. The second school sees factions as the political manifestation of social relationships (Chao 1997; Kao 2004). These relationships can be based on a variety of relationships, including blood, marriage, language, local prox- imity, religious, education, occupation, or other ties. These ties are col- lectively known as "guanxi," which is individualistic, specific, and nonideo- logical (Jacobs 1979, 81). In this relationship of mutual loyalty, there is no master or servant. Rather guanxi relationships reflect interpersonal trust and define the lines of friendships or enmity (Kao 2000, 57). These social networks extend to different parts of the political arena. Chen Chieh- Hsuan sees three distinct levels arranged in concentric circles. Core faction networks form among those most clearly identified with and participating in local politics. They, however, cannot mobilize followers without a wider network of electoral cadres, who are often motivated by symbolic and eco- nomic factors in addition to naked politics. Election cadres, in turn, mobi- lize votes from a civil network (C. Chen 1997, 3 3-48).Jacobs also notes the role of guanxi in mobilizing votes. He observes that voters tend to vote for 106 Mixed-Member Electoral Systems in Constitutional Context candidates who are closest to them. Indeed, guanxi even underlies practices such as vote-buying (Jacobs 1979). The vision of factions as an expression of social relationships implies limits and challenges to the factions. If a faction grows too large, the rela- tionships that undergird it may be weaker. There is a limit to the number of intense social relationships that any single person can maintain, and a suc- cessful faction almost necessarily has to be based in a specific geographic location to ensure sufficient density of these guanxi networks. Moreover, any societal change that shakes guanxi networks is a direct threat to the survival of local factions. For example, economic development, deepen- ing urbanization, the passing of the traditional agricultural lifestyle, and a generational shift from one set of faction bosses to the next all constitute serious challenges to faction structures built on interpersonal relationships (Huang 2010, 232-33; Wang 2004). Bifactionalism All local factions have been confined to a single county. This is partially due to the limitations imposed by social networks. In addition, the KMT did not want any political force to grow large enough to challenge its rule. As the top-tier patron, it could and did cut off the lifeline of patronage bene- fits flowing to any politician who tried to expand his network across county lines. Moreover, there was little incentive for faction politicians to try to construct wider networks. During the authoritarian era, the highest level positions open to elections were all elected at the county level. There were no elections for provincial governor, much less president, so no politician ever had a need to appeal to the national electorate. Some local politicians were recruited into the central government, but these promotions were always based on the politician's personal performance and never on his factional identity. Once a faction politician was promoted into the central government, he was expected to leave his factional priorities behind and work for the KMT regime. Central-local faction linkages were not allowed during the authoritarian era. If the KMT was careful to limit factions' power by limiting their geo- graphic scope, it also limited their power by preventing them from monop- olizing power within that area. In almost all counties, the KMT cultivated a bifactional system and ensured the stability of this system with a rotation of power policy. After one faction had held a particular post for two terms, the KMT nominated a politician from the other faction. Moreover, dif- ferent posts were distributed among the two factions. The key position is Executive Competition, Electoral Rules, and Faction Systems in Taiwan 107 the county magistrate, which controls far more resources than the other positions. In Taiwan's centralized political system, the county assemblies have very little legislative power. Assemblies can supervise the county bud- get, but they cannot demand increases in spending. Assembly members have access to a small pool of resources through each county's Small Pub- lic Works Projects Coordinated Fund, but all expenditures from this fund require the assent of the county magistrate.4 Overall, the county magistrate had more resources at hand and more leeway in distributing them than any other local officeholder. As a result, factions wanted first and foremost to win control of the county government. The KMT was always careful to make sure that neither faction could hold this post for more than two four-year terms. The bifactional system with rotation of power benefitted the factions by preventing the rise of new factions, giving them periodic access to the county magistrate, and ensuring a steady, if smaller, stream of resources even when they were out of power. It benefitted the KMT because the factions checked each other's power and no single faction could stay in power long enough to achieve absolute local dominance (Jen 2006). Taichung County is a nice example of this bifactional system. Prior to its incorporation into Taichung City in 2010, Taichung County's politics had been dominated by the Red and Black factions since the beginning of the ROC era.5 In the 1950s, the two major competitors for county magis- trate, Lin He-nian and Chen Shui-tan, organized the Red and Black fac- tions, respectively, to support their political ambitions.6 In some counties, faction systems reorganize when founders fade from active politics. For example, each new county magistrate in Yunlin County tends to build a new faction. In other counties, like Taichung, the old factions are durable enough to survive a leadership transition.7 In table 4.1, we can see that neither the Red faction nor the Black fac- tion ever held the county magistrate for more than two terms. The faction out of power usually held the speaker's chair. In fact, the KMT tried to give the faction out of power the speaker's seat regardless of whether or not that faction had more assembly members than the other faction. During the authoritarian era, the KMT's dominant position made it unwise for fac- tions to challenge the system. The most notable failure came in 1997, after democratization had weakened the KMT's absolute control. The Black faction magistrate had somewhat neglected grassroots operations during his two terms in the 1980s. The Red faction further starved Black faction politicians while they held power over the next two terms. By 1997, the Black faction had been significantly weakened, and the Red Faction did not want to yield power. Both factions ran a candidate for the county magis- 108 Mixed-Member Electoral Systems in Constitutional Context trate and threw the race to the DPP. Chastened by four years out of power, the two factions cooperated in 2001 to elect a Black Faction politician. Interestingly, the bifactional system was challenged during this period by a third contender, the Yang faction. The Yang faction was led by busi- ness tycoon Yang Tien-sheng, who had risen as part of the Red faction. However, Yang was always on the periphery of the Red faction, and in the early 1990s he and his allies (who came from both major factions) began to call themselves the Third Force. The Yang faction reached its high point in 1994 when his son was elected vice-speaker of the Provincial Assem- bly. However, in 1998 the Provincial Assembly was abolished, and Yang's son moved over to the legislature where he was just one of 225 members. Moreover, in making this move, he displaced the Yang faction's other major source of power, a two-term incumbent who had to yield his seat. The Yang faction lost this seat in 2004, and it subsequently faded away as a political force. When the Black faction waned in strength in the 1990s, the KMT stepped in to nominate a Black faction member for county magis- trate in 1997 and 2001 and thus ensure the continuing viability of the fac- tion. In contrast, the KMT did nothing to help the Yang faction survive.8 TABLE 4.1. Rotation of Power between Factions in Taichung County Taichung County Magistrate Taichung County Council Year Magistrate Party Faction Year Speaker Party Faction 1951 Lin He-nian KMT Red 1951 Li Chen-chung KMT Red 1954 Chen Shui-tan KMT Black 1953 Li Chen-chung KMT Red 1956 Liao Wu-hu KMT Black 1955 Chiu Hsiu-sung KMT Black 1957 Lin He-nian KMT Red 1957 Wang Ti KMT Red 1960 Ho Chin-sheng KMT Black 1961 Wang Tzu-kuei KMT Black 1964 Lin He-nian KMT Red 1964 Wang Tzu-kuei KMT Black 1968 Wang Tzu-kuei KMT Black 1968 Tsai Chiang-yin KMT Red 1972 Chen Meng-ling' KMT Red 1973 Hsieh Yu-ho KMT Black 1978 Chen Meng-linga KMT Red 1978 Hsieh Yu-ho KMT Black 1981 Chen Keng-chin KMT Black 1982 Huang Cheng-yi KMT Red 1985 Chen Keng-chin KMT Black 1986 Huang Cheng-yi KMT Red 1989 Liao Liao-yi KMT Red 1990 Lin Min-lin KMT Red 1993 Liao Liao-yi KMT Red 1994 Lin Min-lin KMT Red 1997 Liao Yung-lai DPP New Tide 1998 Yen Chin-piao IND Black 2001 Huang Chung-sheng KMT Black 2002 Chang Ching-tang KMT Red 2005 Huang Chung-sheng KMT Black 2005 Chang Ching-tang KMT Red 2010 Jason C. Hub KMT none 2010 Chang Ching-tang KMT Red 'Chen Meng-Ling was nominated in the first stage of the KMT's faction replacement policy (Chen 1995, 185). But after he was elected, he cooperated closely with the Red faction. We here mark him as Red faction. bJason C. Hu had been the mayor of Taichung City from 2001 to 2010 prior to the merger of Taichung County and City in 2010. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 109 The faction never was able to secure the resources of a township govern- ment or farmers association, much less the county government. Even with the financial backing of Yang's corporation, it simply could not maintain a viable third faction with the resources from one legislative seat. The Yang faction example is a vivid illustration of just how high the barriers to entry were in the bifactional system. SNTV Elections and Faction Competition A variety of assemblies have elected seats under SNTV rules in Taiwan, including the Legislative Yuan, National Assembly, Provincial Assembly, county assemblies, and township councils. Factions competed for all of these posts, though in the democratic era the legislative and county assem- bly seats have been the most important positions. These offices are criti- cal to a faction's ability to commandeer resources from various levels of government and provide constituency service to supporters, especially for the faction out of power. Factions also use these elections to maintain their mobilization machinery. However, the SNTV elections are not as crucial as those for the county magistrate. With multiple seats available, the KMT could choose from several dif- ferent nomination strategies depending on the strength of the local fac- tions and the strength of other parties. In districts where the factions were strong and the KMT's grip was firm, the KMT might nominate a full slate of candidates, one for each seat available. The nominations would be apportioned to each faction according to their relative strength. In other districts, the KMT might leave one or two seats open for friendly independents who often were faction members. In those districts in which the opposition parties had more support than the KMT, the KMT might nominate candidates for only half the seats and then also support friendly independents (C. Wu 2002, 89). This strategy aimed at "sandwiching" the anti-KMT politicians, with various candidates each taking a bite out of the anti-KMT candidates' support. In some contests in which the fac- tions could not internally determine who should receive a nomination, the KMT would open the race up so that the contestants could have a free fight. However, these free fights were generally carried out when the fac- tion could be sure that one of the two would win; factions were wary of allowing free fights that might harm the faction's overall interests. Overall, the KMT frequently nominated multiple candidates from the same fac- tion in a particular district. To put it another way, the county magistrate elections shape the bifactional system; the two factions then nominate the 110 Mixed-Member Electoral Systems in Constitutional Context best they can in SNTV elections. This is quite unlike the Japanese LDP's factional system, which has taken shape according to the logic of SNTV competition and the need to control a majority in the national legislature in order to form a government.9 The two factions in Taichung County usually nominated multiple can- didates in county assembly elections. Table 4.2a shows nominees and seats won for each faction in 2005 and 2010. If we look only at the eight dis- tricts with two or more seats, the Black faction ran multiple candidates in nine of the sixteen races while the Red faction was represented by multiple candidates in all sixteen races. In fact, it was not unusual for a faction to have three or four candidates in a single race. Table 4.2b shows the party affiliations of the candidates in 2005. From this table, we can see that the factions are affiliated with but also a bit separated from the KNIT. All 29 of the KMT winners in 2005 belonged to one of the factions. However, there were also 10 other faction members who ran and won as independents. Table 4.2c shows that the same pattern appears in 2010. Ten of the 13 KMT winners were faction members, and nine of the 19 winning faction mem- bers ran as independents in 2010. Some of these ran with the tacit approval of the KMT, but others were not nominated and ran on their own. During the democratic era, faction members have become increasingly likely to defy party orders. One specific manifestation of this is that the KMT has a harder and harder time controlling the number of candidates competing for the Pan-Blue pool of voters. The local KMT is still dominated by the factions and the factions are still associated with the KMT, but the KMT can no longer demand absolute obedience to party discipline. Table 4.3 shows the relevant data for legislative elections. Taichung County was a single election district, electing 11 seats in the three SNTV elections in 1998, 2001, and 2004. In these, we also find that each faction regularly nominated multiple candidates and won multiple seats. The Red faction ran four candidates in each of the three elections, while the Black faction had five, six, and three. Seventeen of the 26 candidates from the two big factions were KMT members, but four were independents, two ran (unsuccessfully) under the banner of the People First Party, and one Black faction member even went over to the Green camp, running (and losing) twice under the Taiwan Solidarity Union label. If the number of candidates indicates that factions were not organized to nominate optimally for SNTV elections, there is also another way in which the Taichung experience demonstrates the primacy of competition over the local executive seat. The Red faction has arguably been Taiwan's most successful faction in terms of producing speakers. A Red faction mem- TABLE 4.2A. 2005 and 2010 Taichung County Assembly Election Results by Faction Affiliation Total Seats Black faction Red faction Yang faction 2005 2010 2005 2010 2005 2010 2005 2010 District Seats Seats Seats won Seats won Seats won Seats won Seats won Seats won 2005 (2010) (candidates) (candidates) (candidates) (candidates) (candidates) (candidates) (candidates) (candidates) District 4 (1) 5 (11) 3 (9) 3 (3) 0(3) 2 (2) District 5 (2) 7 (13) 5 (10) 3 (4) 1 (5) 3 (3) District 6 (3) 7 (13) 5 (11) 3 (3) 2 (5) 3 (3) District 1 (4) 8 (17) 5 (10) 1 (2) 0 (2) 2 (2) District 2 (5) 9(18) 6(11) 1 (1) 1 (2) 5 (5) District 8 (12) 6 (13) 4(7) 1 (1) 0(1) 3 (3) District 7 (13) 9 (20) 6 (16) 1 (1) 1 (1) 4 (4) District 3 (14) 4 (7) 2 (6) 1 (1) 0 (1) 1 (3) District 9 (15) 1 (2) 1 (3) 0(0) 0(0) 0(0) District 10 (16) 1 (3) 1 (3) 1 (1) 0 (1) 0 (0) 2 (3) 3 (3) 2 (4) 1 (3) 2 (4) 2 (3) 1 (4) 1 (3) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0(0) 0(0) 0(0) 0(0) 0(0) 0(0) 0(0) 0(1) 1 (1) 0(2) 0(0) 0(0) 0(0) 0(0) 0(0) 0(0) Total Sum. 57 (117) 38(86) 15 (17) 5 (21) 23 (25) 14(27) 1 (1) 0(3) % winners 88.2% 23.8% 92.0% 51.9% 100.0% 0.0% Note: The 2005 Taichung County Assembly had 57 seats elected from 10 districts. The 2010 Taichung Municipal Council had 63 seats elected from 16 districts. Of these, 38 were in the former Taichung County, and 25 were in the former Taichung City. The district boundaries were unchanged in 2010, but the districts were renumbered and each district was apportioned fewer seats. The local factions held 68.42% (39/57) of the seats prior to the merger and 30.16% (19/63) of the total seats, or 50% of the seats from the former Taichung County, after it. Blank cells indicate that no candidates ran. Districts in the former Taichung City are not shown. 112 Mixed-Member Electoral Systems in Constitutional Context ber, Tsai Hung-wen, was speaker of the Provincial Assembly from 1972 to 1981, when that position was commonly considered the highest posi- tion open to elected politicians. His protege, Liu Sung-pan, was speaker of the Legislative Yuan from 1992 to 1999, by which time the legislature had surpassed the Provincial Assembly in importance. With such prominent figures, if any local faction were to reorganize itself to support a leader running in SNTV elections, it arguably should have been the Red faction. However, even Tsai and Liu had to contend with other Red faction nomi- nees.10 Tsai and Liu had tremendous national prestige, but the Red faction county magistrates of the 1970s and 1990s, with their control of the local treasury, were arguably the real faction leaders. The Effect of Electoral and Administrative Reform In 2008, the MMM system was used for the first time in legislative elec- tions. This proved to have very little effect on the KMT's local factions. On the one hand, we have already seen that the KMT's factions are organized around county magistrate elections, not legislative elections. Further, only the legislature used the new system. The county assembly and township council elections continued to be elected under SNTV rules. On the other TABLE 4.2B. 2005 Taichung County Assembly Election Results by Party Affiliation KMT DPP PFP TSU All others Total seats Seats won Seats won Seats won Seats won Seats won District (candidates) (candidates) (candidates) (candidates) (candidates) (candidates) District 4 5(11) 3(3)R2B1 0(2) 2(6)B2 District 5 7(13) 3(5)R2B1 0(2) 0(1) 4(5)R1B2 District 6 7(13) 5(6)R2B3 1(2) 1(5)R1 District 1 8(17) 3(5)R2B1 3(3) 1(1) 1(1) 0(7) District 2 9(18) 5(8)R4B1 3(4) 0(1) 0(1) 1(4)R1 District 8 6(13) 3(4)R2B1 2(3) 0(1) 0(1) 1(4)R1 District 7 9(20) 5(6)R4Y1 2(4) 1(1) 0(1) 1(8)B1 District 3 4(7) 2(4)R1B1 1(1) 0(1) 1(1) District 9 1(2) 0(1) 1(1) District 10 1(3) 1(3)B1 Total 57(117) 29(42) 12(21) 3(6) 1(5) 12(43) R19B9Y1 R4B6 % winners 69.0% 57.1% 50.0% 20.0% 27.9% Seats share 50.9% 21.0% 5.3% 1.8% 21.0% Notes: abbreviations = B: Black faction; R: Red faction; Y: former Yang faction. The table should be read as follows: in District 4, "3(3)R2B1" means the KMT nominated 4 candidates and won 3 seats; of the 3 winners, 2 belong to the Red faction and 1 belongs to the Black faction. "0(2)" means the DPP nominated 2 candidates and won no seats. "2(6)B2" means 6 candidates ran with no party affiliation, and 2 of them won. More important, the winners both belong to the Black faction. Blank cells indicate that no candidates ran. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 113 hand, the single-seat districts available to the local factions under MNFM should push toward bifactionalism." Since the KMT already had a bifac- tional system in most counties, it would be strange to expect the new elec- toral system to bring about significant change. Taichung County was apportioned five seats under the new system. Instead of competing for each nomination, the factions decided to con- tinue the collaboration they had practiced since the 2001 county mag- istrate election. The six faction incumbents divided up the five districts among themselves. The KMT cemented this bargain by giving the sixth incumbent a place on the KMT party list and by agreeing not to nominate a candidate against the Black faction incumbent running as an independent in District 2. The candidates even traded support across district lines by exchanging election cadres (Huang, Wang, and Wang 2009). This strategy was a complete success, as the factions won all five districts. In a certain sense, the new electoral system seemed to strengthen the local factions by placing all of Taichung County's legislative seats in their hands. In 2012, the DPP managed to take two of the five seats, but the blame for these defeats should go primarily to the KMT, not to the factions. In District 1 the Red faction incumbent retired, and the factions were unable to come to a consensus on a new candidate. A local high school principal loosely associated with the Black faction was the only person to register for the TABLE 4.2C. 2010 Taichung Municipal Council Election Results by Party Affiliation KMT DPP PFP TSU All others Total Seats Seats Won Seats Won Seats Won Seats Won Seats Won District (candidates) (candidates) (candidates) (candidates) (candidates) (candidates) District 1 3(9) 1(2)R1 1(1) 1(6)R1 District 2 5(10) 1(2)B1 1(2) 3(6)R3 District 3 5(11) 3(3)R1B2 1(2) 1(6)R1 District 4 5(10) 1(4)R1 2(3) 1(1) 1(2) District 5 6(11) 3(4)R2 2(3) 1(4)B1 District 12 4(7) 1(3)R1 2(2) 1(2)R1 District 13 6(16) 0(4) 3(3) 1(1) 2(8)R1B1 District 14 2(6) 1(2)R1 1(1) 0(3) District 15 1(3) 1(3) District 16 1(3) 1(3) Total 38(86) 13(30) 13(17) 1(1) 1(1) 10(37) R7B3 R7B2 % winners 36.4% 76.5% 100.0% 100.0% 27.0% Seats share 34.2% 34.2% 2.6% 2.6% 26.3% Note: For instructions on how to read this table, please refer to the note in table 4.2b. In the full Taichung Municipal Council (including the former Taichung City), the KMT won 27 seats, the PFP won 1 seat, the DPP won 24 seats, the TSU won 1 seat, and independents won 10 seats. 114 Mixed-Member Electoral Systems in Constitutional Context KMT's nomination. The factions did not take him seriously and continued to negotiate, but to everyone's surprise the principal managed to pass the threshold of approval in the KMT's noncompetitive telephone surveys" and was duly nominated. The DPP thrashed this very weak nominee. In District 7 (formerly District 3), the 2008 winner was convicted of vote buy- ing and stripped of his seat. The DPP then won the seat in a by-election. In 2012, the KMT chose to parachute in a candidate from Taipei with no local connections. She was beaten soundly by a DPP candidate with deep local roots. In the other three districts, however, the factions proved their continuing strength. Overall, factions have adapted to the new legislative electoral system TABLE 4.3. Taichung County Legislative Election Results, 1998-2012 Year (Blue : Green) KMT DPP New PFP TSU Others 1998 (9:2) Nominees 8(R3,B4,Y1) 4(N1,NDl,nf2) 1 R1 Seats won 7(R3,B3,Y1) 2(N1,ND1) 1 R1 2001 (7:4) Nominees 6(R3,B2,Y1) 5(N3,ND1,NE1) 1 4(R1,B1,nf2)1(B1) B2 Seats won 5(R3,B1,Y1) 5(N3,ND1) 0 1(nfl) 0 B1 2004 (7:4) Nominees 5(R4,B1) 6(N3,W1,GFA1,nfl) 1(nfl) 2(B1,nfl) B1 Seats won 5(R4,B1) 4(N2,W1,nfl) 1(nfl) 0 B1 2008 (5:0) District 1 1(R1) @ 1(N1) District 2 0 1(B1) @ District 3 1(R1) @ 1(N1) District4 1(R1) @ 0 1(nfl) District 5 1(R1) @ 1(N1) 2012 (3:2) District 1 1(B1) 1(N1) @ 1(nfl) District 2 1(nfl) 1(B1) @ District 7(3) 1(nfl) 1(nfl) @ 1(nfl) 1(nfl) District 8(4) 1(R1) @ 1(N1) 1(nf) 2(B1,nfl) District 3(5) 1(R1) @ 1(nfl) Na 1(F1) Notes: abbreviations for KMT factions are B: Black faction; R: Red faction; Y: former Yang faction; abbrevia- tions for DPP factions are N: New Tide; W: Welfare; ND: New Dynamic; NE: New Era; GFA: Green Friend- ship Alliance; for Others, nf: no faction affiliation. Blank cells indicate no candidate nominated. The table should be read as follows: the cell "8(R3,B4,Y1)" means the KMT nominated 8 candidates, includ- ing 3 Red faction members, 4 Black faction members, and 1 Yang faction member. Under SNTV, Taichung County was one electoral district and elected 11 seats. After the shift to MMM in 2008, Taichung County was divided into 5 single-seat districts. The mark "0" shows the winning candidate in MMM elections. The district boundaries were not altered in 2012, but they were renumbered. For example "District 7(3)" indicates that the 2012 District 7 was District 3 in 2008. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 115 with a minimum of disruption. Administrative reform is a much more seri- ous threat to the KMT's local factions. In 2010, Taichung City and County were combined to form Taichung Municipality, one of four new direct municipalities."' Direct municipalities differ from counties and county-level cities in a number of ways. Munici- pality mayors have much larger budgets with much more discretionary spending, and they can appoint many more positions within the munici- pal government. Counties have elected township mayors and councils. In direct municipalities, townships are converted into administrative districts, and the executives and councils are appointed'4 and politically irrelevant. From the perspective of local factions, there were far fewer offices avail- able," and, to win the biggest prize, the municipal mayor's seat, they had to compete not only with each other and the opposition party but also with all the politicians and factions from the other city or county. Since the only useful positions left, other than the five legislative seats, were in the municipal council, the 2010 election was particularly intense. With fewer seats available, many incumbents were unable to secure a nomination from the KMT. Rather than accept the death of their politi- cal careers, many faction members ran as independents. The result was a bloodbath. The KMT did terribly, with only 13 of its 30 nominees win- ning. The Red faction had a miserable performance, as only 14 of its 27 candidates won. The Black faction was devastated; only 5 of its 21 members won. To put this in perspective, in 2005 the two factions had elected 38 of their 42 candidates. The main reason for the electoral disaster of 2010 was that the factions simply had too many candidates running. Another reason is that they now needed to win more votes. Compared to 2005, success- ful candidates in 2010 needed 70-100% more votes. Many politicians had built their machines focusing on particular areas or constituencies to fit the old system. Their networks were the right size to deliver the old number of votes, but when factions are built on personal relationships, it is no small task to scale up to a meet the new, higher threshold. With no township governments and without control of the municipal government, both factions are left without direct control over any local government resources. They are not entirely powerless, of course. Since the mayor came from Taichung City, the KMT helped Taichung County win control of the municipal council. The new speaker is from the Red faction, while the vice-speaker is from the Black faction. Moreover, while the two factions have less than one-third of the total seats, this should be sufficient for them to guide the council's agenda if they cooperate. Within the KMT, the two factions have 10 of 27 seats, but if one looks at the 116 Mixed-Member Electoral Systems in Constitutional Context broader Blue camp, the two factions combine for 19 of 38 seats. However, it remains to be seen if the factions will try to cooperate with each other, look for allies from the former Taichung City, or simply fade from promi- nence as they are starved of patronage resources. KMT Local Factions: Summary In the KMT's authoritarian regime, elected politicians were effectively locked out of competition for national power. In a sense, whether the national executive was presidential or parliamentary was irrelevant to KMT local factions. Instead, factions organized around the most desirable post within their grasp, in this case local executive office. As these posts were elected by the plurality rule, competitors had to form preelectoral coalitions, and these formed the basis of two stable factions. The KMT reinforced this structure by actively intervening to ensure that two, and only two, factions remained viable. These factors were sufficient to main- tain bifactionalism even when all assembly seats were elected under SNTV rules. The shift to MMM in legislative elections does little to change this calculus. In fact, the factions have collaborated more often than they have competed, yielding seats to one another instead of directly fighting one another for each seat. Looking forward, the factor with the greatest poten- tial to shake up the system is administrative reform, which has deprived the local factions of most of their patronage resources. DPP National Factions In contrast to the KMT's factional structure, the DPP's internal factions are structured more similarly to those of the Japanese LDP. Throughout the democratic era, the DPP has usually had around four major factions, all of them national in scope. Until 2006, the factions were formal organiza- tions, with bylaws, regular meetings, and membership rosters. In 2006, the DPP formally abolished its factions, and the factions have been informally organized since then. The DPP's factional system arose in the era before the presidency became a reasonable goal. When the presidency did eventually come into play, it placed considerable pressure on the factional system to evolve. Indeed, most of the major upheavals of the past 15 years have been directly related to presidential politics, as factions have allied and reshuffled to try Executive Competition, Electoral Rules, and Faction Systems in Taiwan 117 to win the presidential nomination and risen or fallen with their leaders' prospects. Arenas of Competition While KMT factions are designed to capture and distribute patronage resources at the county level, the major goal of DPP factions is to help their members gain control of the national party machinery and national elected offices. While the presidency has become a focal point of competi- tion in recent years, factions originally competed in three primary are- nas. Even with the increasing focus on the presidency, these three arenas remain important in shaping the DPP factional system. First, they compete for nominations to elected office. Like the KMT's local factions, DPP fac- tions try to win local executive offices. However, since they are national in scope, there is not simply one big prize. DPP factions can win some nominations while losing others without facing an existential crisis. In fact, DPP factions place a higher priority on winning large numbers of legisla- tive seats, and the size of a faction's legislative delegation is the clearest measure of a faction's strength and influence. Factions can help members win nominations and elections by providing financial or human resources, coordinating among aspiring candidates, and mobilizing party member or regular voters in primaries. The DPP's nomination rules have changed considerably over the years (see chapter 5), but factions have always played a central role in deciding nominations. Second, the factions compete for positions within and control of the party machinery. The most important of these positions are seats in the Central Standing Committee (CSC), the body that makes most of the criti- cal decisions. The CSC is indirectly elected. Party members elect represen- tatives to the National Party Congress, which meets only once a year. The National Party Congress elects 30 of its members to the Central Executive Committee (CEC), and the Central Executive Committee elects 10 of its members to the CSC.16 All of these elections are conducted by SNTV. Needless to say, a well-organized group can do quite well in this three- stage indirect election. Indeed, factions often house their members in spe- cific hotels for the weekend, with each faction based in a different hotel. This allows the faction leaders to give members voting instructions and, more important, to keep them away from other factions who might seek to poach their votes. Factions try to maximize their power by enrolling party members, directing them how to vote, and using extra votes to support 118 Mixed-Member Electoral Systems in Constitutional Context allies. No faction has held a majority of seats in the CSC since the earliest years of the party. Instead, the factions form postelectoral coalitions within the CSC. Until 1996, these coalitions elected the party chair. The chair has been directly elected since 1998, and even though this yields some inde- pendent stature, party chairs have had to come to an accommodation with a majority in the CSC in order to lead the party effectively. The third and least important arena of competition is for the three floor leaders of the DPP legislative caucus. The DPP elects new floor leaders for each six-month session, and these elections have generally been contested along factional lines. Again, these elections use SNTV, so it is not necessary to have a majority to win a caucus leader position. Prior to electoral reform, nominations to the legislature, elections to the CSC, and floor leader elections were all conducted under SNTV. With multiple seats available, multiple factions could survive. After electoral reform, CSC and floor leader elections continue to be held under SNTV. Moreover, the CSC remains an arena in which coalitional, not majori- tarian, politics hold sway. In addition, multiple factions can compete for two or three legislative seats each in the party list tier. While the electoral reform has eliminated one of the major institutional foundations of the multifactional system, it has not eliminated all of them. Characteristics of the Individual DPP Factions When the DPP was founded in 1986, it had two major factions: Formosa and New Tide. Formosa members tended to be bound together by per- sonal, rather than ideological, ties, and they tended to be quite pragmatic, interested first and foremost in winning office (Rigger 2001, 72; Cheng and Hsu 1996). With only one exception, all other DPP factions also adopted a similarly pragmatic orientation. DPP factions are not noted for clear positions on specific issues or distinct ideological positions. Factions tend to focus on maximizing power and leave the policy decisions to indi- vidual officeholders. Perhaps because the glue holding them together is relatively weak, the factions tend to have loose discipline. Members usually vote with their faction, but often they do not. When they defect, there are few repercussions. Because of this weak discipline, faction leaders have weak bargaining power. The leader may make a promise, but he cannot always commit the full strength of his faction to that promise. The New Tide faction is the exception. New Tide is highly organized and highly disciplined. It is organized along Leninist lines and considers itself a vanguard organization (Cheng 2004, 149). Members are carefully Executive Competition, Electoral Rules, and Faction Systems in Taiwan 119 screened before being accepted into the faction, and they participate in regular training and education meetings. During elections, the faction pro- vides campaign staff and funds for its candidates. In the 1980s and 1990s, New Tide was considered to have clear ideological stances favoring Taiwan independence and social welfare, though these distinctions have become murkier over the last decade. New Tide members also take militant posi- tions against corruption. The faction practices democratic centralism, and all members are expected to obey faction decisions. Violators are almost always expelled from the faction. New Tide's discipline allows it to exert much more influence than its numbers imply. Faction leaders are cred- ible negotiators, since counterparts know they can deliver on their prom- ises. Moreover, New Tide can make standing commitments of support not just today on a particular question but also in the future on unspecified questions. This makes New Tide particularly attractive as a coalition part- ner. However, New Tide's discipline and disproportional influence arouse powerful emotions among other DPP members. Other factions can be jealous of New Tide's power, they grumble about New Tide's voting bri- gades, and they have at times worried that New Tide's militant political positions might hurt the DPP at the ballot box. No other faction arouses such passions, and New Tide always has to worry about the potential of an "everyone-against-New Tide" coalition. The Welfare State Alliance and Justice Alliance formed in the early 1990s. With democratization, all the seats in the legislature were up for election, creating space for more than two factions.Justice was centered on Chen Shui-bian and included a few of his political allies. In the early 1990s, Justice was the smallest of the four main factions. Welfare was nominally led by Frank Hsieh, but it was more accurately understood as a coalition of everyone not in Formosa, New Tide, or Justice. Welfare had several promi- nent leaders, and they were not likely to follow orders from each other.'7 A Brief History of DPP Factional Politics in the Presidential Election Era In this subsection, we present a rough history of DPP factional politics over the past decade and a half. During this period, the DPP's factions underwent substantial changes, and most of these changes can be traced, at least in part, to conflicts over the presidency. If the 2000 presidential election was the first one that the DPP had a chance of winning, then the first major change to the factional system in the era of competitive presidential politics was the splintering of the For- 120 Mixed-Member Electoral Systems in Constitutional Context mosa faction. Formosa's leader and presidential aspirant, Hsu Hsin-liang, argued that if the DPP wanted to win power, it should be more open to economic integration with China. This stance put many Formosa mem- bers in an uncomfortable position with their supporters (UDN, December 20, 1998). Taking advantage of dissatisfaction with Hsu's leadership, his top lieutenant, Chang Chun-hung, usurped control of the faction. By the July 1998 CEC and CSC elections, the Chang faction and the Hsu faction had effectively split, with each organizing its own vote rationing scheme. Most of the Formosa faction ended up in Chang's new group, New Era.'8 Hsu's supporters renamed themselves New Dynamic. Already small, New Dynamic was further crippled by a lack of leadership, as Hsu withdrew from the DPP to focus on his unsuccessful presidential bid (UDN, June 7, 1999). New Era and New Dynamic continued to exist until about 2004, but neither was able to exert much influence in the party. With no charismatic leaders and no patronage to offer, both slowly dwindled in size. Around 2004, a small new faction, the Green Friendship Alliance, absorbed most of the remnants of New Era and New Dynamic. However, in terms of influ- ence, the Green Friendship Alliance was merely an echo of the old For- mosa faction. The disintegration of Formosa was a direct result of Hsu's unsuccessful pursuit of the presidency. Chen Shui-bian won the presidency in March 2000, and his faction, Jus- tice, grew quickly (see tables 4.4 and 4.5). In the past, pragmatic politicians who wanted to be on the winning side might have joined Formosa; now they joined Justice. Justice moved quickly to try to put together an even bigger coalition. In April and May, leaders of the Justice, Welfare, New Dynamic, and Taiwan Independence Alliance factions had meetings to dis- cuss cooperation and even merger. On May 20, they announced they would form the Mainstream Coalition, which included around 40 legislators from all factions except New Era and New Tide (UDN, April 19, 2000; UDN, May 5, 2000; UDN, May 13, 2000; UDN, May 20, 2000). This Mainstream Coalition certainly looked like the first step in reorienting the system into a bifactional system, but it was not to be. The Mainstream Coalition never expanded its activities from within the legislative caucus to other arenas, such as competition for nominations or CSC seats. Moreover, many of the members, especially those from Justice, quietly withdrew from Mainstream Coalition activities, and the coalition quickly became a small, marginalized group dominated by Welfare members. Instead of Welfare, President Chen chose the larger and more reliable New Tide as the main coalition partner for his Justice faction. Chen put several New Tide leaders into important positions in the executive branch. TABLE 4.4. Size of Each DPP Faction in the Legislative Yuan 1995 1998 2001 2004 2008 2012 Formosa New Tide Welfare Justice TIA New Era New Dynamic GFA Su Hsieh Chen Yu Grandparents No faction 14 11 13 11 4 15 12 13 5 19 15 34 4 22 12 20 1 4 7 14 5 6 4 5 1 1 1 3 5 4 5 3 1 2 1 1 2 4 6 29 9 19 Total DPP delegation 55 69 87 89 27 40 Note: Factional affiliations for 1995, 1998, and 2001 are taken from Cheng (2004, 563-76). Fac- tional affiliations for 2004, 2008, and 2012 were compiled by the authors from newspaper reports. TABLE 4.5. Size of Each DPP Faction in the Central Standing Committee 1994 1996 1998 2000 2002 2004 2006 2008 2010 Formosa New Tide Welfare Justice TIA New Era New Dynamic GFA Su Hsieh Chen Yu Grandparents No faction/other 4 3 2 3 2 2 3 2 2 2 3 3 3 2 4 3 1 1 1 2 3 3 2 1 2 1 1 1 2 1 2 2 2 1 1 1 1 1 1 2 1 1 1 2 0 1 1 1 1 1 1 1 2 1I Total CSC seats 11 11 11 11 10 10 10 10 10 Note: Factional affiliations for 1994-2002 are taken from Cheng (2004, 214-17). Factional affiliations for 2004-2010 were compiled by the authors from newspaper reports. Only directly elected seats are in- cluded; seats automatically awarded to mayors and legislative floor leaders are not included in this table. 122 Mixed-Member Electoral Systems in Constitutional Context In fact, because Justice members were relatively junior and not yet qualified for such important posts, New Tide was more successful than Chen's own Justice faction at placing its members into the government bureaucracy. The Chen-Justice-New Tide axis dominated the DPP throughout Chen's first term. While Justice had more members, its internal discipline was dis- mal, and Chen often had to rely on New Tide to get what he wanted. Chen's choice of coalition partners was likely the most important rea- son that the DPP maintained a multifactional system. If Chen had sup- ported the Mainstream Coalition and encouraged its various members to merge into one new faction, he likely would have been able to reshape the system into a bifactional Mainstream vs. New Tide system. Instead, he chose to cooperate with New Tide. The very internal cohesiveness that made New Tide such an attractive coalition partner also ensured that it would remain a separate entity outside of Chen's direct control. Moreover, because New Tide retained its independence, it had the capacity to termi- nate the coalition. New Tide began to distance itself from Chen in the early months of his second term. As public outrage at Chen's mounting corruption scandals grew, New Tide moved into open attacks on President Chen. By June 2006, Chen had been so severely weakened'9 that he was forced to announce he would step back from the first line of politics and devolve power to Premier Su Tseng-chang. New Tide continued to apply pressure. When the DPP declined to ask Chen to resign or to recommend sending his case to the government anticorruption agency, two New Tide leaders resigned their seats in the legislature, an action they claimed was designed to shock the DPP into self-reflection (CT, November 14, 2006). By the end of 2006, New Tide was in open revolt against President Chen. As Chen stumbled toward the end of his tenure, the battle to succeed him as president intensified, and this contest reshaped the DPP's factions. The two leading candidates, Su Tseng-chang and Frank Hsieh, both had roots in the Welfare faction, but Welfare had ceased to be a coherent orga- nization by Chen's second term. Within weeks after Chen's reelection, the media started referring to the Su faction and the Hsieh faction. At first, these were merely a few of Su or Hsieh's closest political allies, and many of them maintained membership in one of the traditional factions. The Su and Hsieh factions would quickly grow into larger groupings, and these factions would contest power in multiple arenas. Unlike the traditional fac- tions, the Hsieh and Su factions were never formal organizations. They might have eventually moved to formalize their networks, but another major shock rocked the DPP's factional ecology. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 12 3 In July 2006, the DPP's National Party Congress voted to dissolve all factions.20 This was not a new proposal; it had been discussed at several previous party congresses but had never won much support. In 2006, how- ever, conditions were ripe for the motion to pass. Once again, many of the tensions were directly related to presidential politics. The motion was an attack on New Tide, which had grown steadily in size and influence over the past decade. Most people in the party felt that New Tide benefitted most from being organized so it would lose the most if it were forced to disband. Moreover, anti-New Tide passions were running high. New Tide had profited from its alliance with President Chen in his first term and had then turned on him. This could not help but infuriate other people in the party who remained loyal to Chen. The motion may also have got- ten a boost from the upcoming presidential race. New Tide had formed an alliance with Su, and Hsieh was beginning to form an anti-New Tide coalition. Because of this, Hsieh's supporters were enthusiastic about the motion not only because it hit directly at New Tide, but also because it indirectly weakened their principal rival for the presidential nomination (CT, July 3, 2006).21 Cheng argues that, in addition to injuring New Tide, the DPP may have had two other reasons for dissolving its factions. First, President Chen may have been trying to avoid blame for his own myriad woes by focusing attention on the DPP's factions. Second, electoral reform had just passed, and the DPP may have felt that multiple factions were no longer needed for legislative elections using single-member districts (Cheng 2010). This latter point is especially relevant to our purposes. However, the electoral system did not play a prominent part in the actual public debates over the merits of dissolving the factions. If factions simply were not useful under the new electoral system, they would simply have faded away. Alternatively, the multiple factions might have simply consolidated into two factions. There would have been no need to actively abolish them. In fact, multiple factions were still useful in contests for CEC, CSC, party list, and city and county council seats. Electoral reform is not a convincing explanation for the dissolution. Factions did not cease to exist simply because the motion passed.22 There had been some worries that New Tide might collectively quit the DPP if such a motion passed, but all factions announced that they would comply with the letter of the new rules. Within a short time, it became apparent that the other factions had been the primary victims of the motion. Without formal offices, rules, and meetings, the other factions had very little to hold them together and quickly withered away. New Tide, in 124 Mixed-Member Electoral Systems in Constitutional Context contrast, was bound together with much stronger glue and continued to operate as a relatively coherent faction. The names of the other factions disappeared almost immediately from media reports following the party congress. In contrast, the media reported regularly on the activities of the "former New Tide" faction.23 With the dissolution of all the traditional factions except for New Tide, most DPP members were suddenly free to form new alliances, and they coalesced around the two major presidential contenders and the incumbent president. After the 2006 party congress, the Su and Hsieh factions grew quickly in size and influence. The Su, Hsieh, and Chen factions were not simply a renaming of the old factions. For one thing, while many of the former Justice members now identified as Chen loyalists, several took this opportunity to switch their allegiance to Su or Hsieh. For another, Su and Hsieh had both originally belonged to the Welfare faction, so there were no obvious cues that might help the old faction members switch en masse to one side or the other. Instead, there was a fairly thorough reshuffling of alle- giances. Because this reshuffling occurred in the context of a fierce battle for the presidential nomination, the new factions consolidated rather quickly. By the end of President Chen's second term in 2008, a new faction sys- tem had taken shape. In addition to the old New Tide faction, there were three other major factions, the Su, Hsieh, and Chen factions. There were also a few smaller factions organized around somewhat less prominent leaders. These new factions were different from the old factions in that they were not formally organized. They did not have formal rosters, rules, or meetings. Membership was not clearly defined, and a particular indi- vidual might have good relations with multiple faction leaders. Moreover, they were much more centered around a single leader than the previous factions had been. Where the old factions had been physically centered in a legislative office and measured by the size of their legislative delegations, the new factions were more focused on presidential contenders than on legislative delegations. However, the new factions continued to organize their members to contest nominations and party offices, just as the old factions had. The factional system has seen relatively minor changes during Ma Ying-jeou's presidency. A new faction may be starting to form around Tsai Ing-wen. Tsai, who was not affiliated with any faction, served as party chair during Ma's first term and eventually became popular enough to win the 2012 presidential nomination. During Ma's first term, she received support from all the major factions. However, after losing the presidential election and resigning as party chair, she established a think tank and has begun Executive Competition, Electoral Rules, and Faction Systems in Taiwan 12 5 cultivating allies in the legislature and the party. If Tsai's influence is rising, former president Chen's is waning. There has been quite a bit of pres- sure for the DPP to draw a clear line between Chen and itself, and the Chen faction has fought to resist this. Chen's faction had some success in the 2010 city council elections, conducted under SNTV rules, but far less in the 2012 single-seat legislative races. Moreover, in the 2010 CEC and CSC elections, the Chen faction was unable to win any CSC seats.24 With Chen's health deteriorating in prison and the battles of Chen's presidency fading from the present, the Chen faction faces a struggle to retain its size and influence within the DPP. Throughout this discussion, one can see again and again how competi- tion over the presidency has molded the DPP's system of factions. The most important events, including the splintering of Formosa, the forma- tion and dissolution of the alliance between Justice and New Tide, the formal abolition of all factions, and the emergence of the Su and Hsieh factions, were all results of competition to control the presidency or efforts by the president to reshape the factional system. Number of Faction Nominees per District Masaru Kohno argues that the number of LDP factions was heavily influ- enced by the SNTV electoral system. As a general rule, factions only nomi- nated one candidate per district. Because the LDP nominated multiple can- didates in almost all districts, this created space for multiple factions. Ideally, a faction would grow large enough to nominate one candidate in every dis- trict nationwide. In practice, they did not grow quite that large, and there were four large factions that nominated in most districts and one smaller faction that nominated candidates only in larger districts (Kohno 1992). Almost all Japanese districts have between three and five seats, so the LDP factions faced a fairly similar challenge across districts. Taiwan's dis- tricts vary much more, ranging from 1 to 17 seats. Moreover, Taiwan's fac- tions have also competed for 30 Central Executive Committee seats, 10 or 11 Central Standing Committee seats, around 15 party list seats, and around 3 Overseas Representatives seats. These very different challenges make it much harder to construct factions of exactly the right size such that a faction can always nominate one and only one candidate. Even so, DPP factions have exhibited tendencies to avoid multiple nom- inees. To illustrate this, we first considered all DPP district nominations worth fighting for in SNTV legislative elections from 1995 to 2004 (table 4.6).5 The DPP nominated 304 candidates in these districts, of whom 256 126 Mixed-Member Electoral Systems in Constitutional Context were affiliated with a faction. Of these faction candidates, 176 (68.8%) did not have to run against another nominee from their faction. However, this overall number hides differences across district sizes. As might be expected, multiple nominees from a single faction were much more common when the DPP nominated more candidates. When the DPP only nominated one or two candidates, 95.3% of the faction nominees did not run against a fel- low faction member. When the DPP nominated three or four, this percent- age fell to only 74.4%. In districts with five or more DPP nominees, half had to face at least one other nominee from their faction. There were also significant differences among the various factions. Multiple nominations were almost exclusively the province of the three biggest factions. Among TABLE 4.6. Number of DPP Faction Nominees in SNTV Legislative Elections Number of faction nominees spots districts 0 1 2 3 4 nominees %solo All factions 1-2 203 161 41 1 43 95.3 3-4 227 125 87 15 117 74.4 5+ 110 44 48 9 6 3 96 50.0 total 540 330 176 25 6 3 256 68.8 New Tide 1-2 37 34 3 3 100.0 3-4 41 9 26 6 38 68.4 5+ 20 3 11 2 4 27 40.7 total 98 46 40 8 4 68 58.8 Welfare 1-2 37 25 12 12 100.0 3-4 41 19 18 4 26 69.2 5+ 20 3 15 2 19 78.9 total 98 47 45 6 57 78.9 Justice 1-2 37 21 15 1 17 88.2 3-4 41 17 20 4 28 71.4 5+ 20 3 7 5 2 3 35 20.0 total 98 41 42 10 2 3 80 52.5 Formosa, TIA 1-2 92 81 11 11 100.0 New Era, New 3-4 104 80 23 1 25 92.0 Dynamic, GFA 5+ 50 35 15 15 100.0 total 246 196 49 1 51 96.1 Unaffiliated 1-2 37 30 6 1 8 75.0 3-4 41 24 10 6 1 25 40.0 5+ 20 9 8 2 1 15 53.3 total 98 63 24 9 2 48 50.0 Note: These data cover the 1995-2004 legislative elections. The following districts are excluded: Taitung County, Penghu County, Jinmen County, Lienchiang County, Plains Aborigines, Mountain Aborigines, Party List, and Overseas Representatives. Factional affiliations for 1995, 1998, and 2001 are taken from Cheng (2004, 563-76). Factional affiliations for 2004 were compiled by the authors from newspaper re- ports. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 127 Formosa, Taiwan Independence Alliance, New Era, New Dynamic, and Green Friendship Alliance, there was only one case of a multiple nomi- nees. However, among Welfare, New Tide, and Justice, 21.1 %, 41.2%, and 47.5% of nominees, respectively, had to run against a fellow faction mem- ber. It is especially striking that in the biggest districts, a majority of New Tide (59.3%) and Justice (80.0%) nominees had to deal with challenges from other faction members. In general, factions usually nominated a single candidate, but the larger factions were not afraid to nominate multiple candidates in larger districts. Even so, there is anecdotal evidence that they viewed nominating mul- tiple candidates as a challenge. For example, in the 2004 election, New Tide originally had multiple candidates in Taipei County 2 and Taichung City. However, four months before the election, it held a meeting in which the extra candidates were encouraged to withdraw from the faction (CT, August 11, 2004). It is probably no coincidence that the two factions that were able to grow large enough to consistently nominate multiple candi- dates and navigate the resultant coordination problems were New Tide and Justice. The former is famous for its internal discipline, and both had ample resources from the executive branch. This last point brings us back to the importance of the executive. If the electoral system created pressures for factions to remain small enough that they would not need to nominate multiple candidates in a district, competition over the executive created both the incentive for bigger factions and the resources to sustain them. DPP National Factions: Summary This section draws attention to important parallels between the DPP and LDP factions. Like the LDP, the DPP has multiple factions competing for control of the national party. Even though the DPP party chair is directly elected, chairs need to forge a working majority in the CSC to be effec- tive. Since no faction has had an outright majority since the very early days of the party, this has required building a coalition of multiple factions. To use the terminology of the introductory chapter, a postelectoral strategy is viable, so there is no need to consolidate into two big factions. Like the LDP, the DPP's factions were born in a SNTV environment, and multi- factionalism helped both parties nominate multiple candidates with differ- entiated power bases. However, neither the LDP's nor the DPP's factions were completely dependent on the SNTV system, and both managed to adapt and survive after electoral reform. If there are similarities between the LDP and DPP factions, there are 128 Mixed-Member Electoral Systems in Constitutional Context also notable differences. The LDP is a party of power, while the DPP has traditionally been an opposition party. The LDP's factions collect and dis- tribute enormous sums of money, and this money is often used by individ- ual MPs to sustain a koenkai (personal support organization). These sorts of resources have generally not been available to the DPP, and collecting and distributing money plays an extremely limited role in the DPP's factional politics. The koenkai help LDP politicians reach out to ordinary voters to win general elections; DPP factions connect politicians with party mem- bers to win primaries. Overall, the theories constructed to understand LDP factions are far more useful for understanding DPP factional politics than for deciphering KNIT factions. Conclusion Our discussion of both KMT and DPP factions points to the centrality of executive competition. However, this was not necessarily competition over the national executive. For KMT local factions and DPP factions until the late 1990s, the presidency was not a viable goal. Rather, the factions orga- nized themselves around the most attainable executive positions. KMT local factions are designed to win control of the local government, and the result is bifactionalism. DPP national factions were originally formed to win control of the national party, and building a minority faction and then constructing a coalition with other minority factions proved a viable strat- egy. When the presidency became a possibility, the DPP's faction system also changed. All the major changes of recent years, including the disin- tegration of the Formosa faction, the push toward a bifactional system in the early years of the Chen presidency, and the evolution from legislative- centered factions to factions centered on presidential contenders, were all intimately related to competition over the presidency. The introduction reminds us that executive competition plays an important role in shaping party and faction systems. This chapter builds on that notion by point- ing out that the critical competition may not always be over the national executive but may instead be over a local executive post, a party leadership post, or some other post that is both desirable and attainable. Another lesson is that the electoral system used for the national leg- islature may be less important than we previously believed. The change from SNTV to MAMFM forced factions to rethink their legislative electoral strategies, but it did not affect them in other critical arenas. After elec- Executive Competition, Electoral Rules, and Faction Systems in Taiwan 129 toral reform, KMT local factions continued to place most of their focus on the single-seat-plurality county magistrate elections. Their secondary arena was arguably in the county assemblies, which continued to be elected by SNTV. Legislative elections were more central to the DPP's nation- ally oriented factions. However, all the DPP's party positions as well as elections for the legislative caucus floor leaders continue to be elected by SNTV. With multiple arenas of competition, it is simplistic to expect that reforming the electoral system in just one arena will completely transform the factional system. Finally, the different paths of the KMT and DPP draw attention to the historical context during the period of faction development. The KMT local factions were created in an authoritarian context in which they were not allowed to compete for national power or espouse controversial politi- cal ideas. The KMT used factions to co-opt local elites and prevent them from coalescing into a force that could threaten its grip on national power. The result was locally oriented factions aimed at distributing patron- age. DPP factions grew up in a very different context. Coming out of the democratization movement, DPP politicians were always focused on national politics. Even if winning power was not a viable short-term goal in the late 1980s and early 1990s, the DPP leaders who formed factions were nationally oriented politicians who forged alliances with similar colleagues from different areas. This led to a party with multiple power centers and a collective leadership. Today's environment is different still from either the authoritarian era or the period of transition to democracy. The fac- tions have evolved to meet current challenges, but the legacy of yesterday remains powerful. NOTES 1. These data are based on a research project subsidized by the National Science Council research program entitled "The Relationship between Local Factions and Political Parities after the Consolidation of Taichung County and Taichung City." Project No. NSC 99-2410-H-029-061. 2. For more detail on the theoretical foundations, historical evolution, and everyday operations of Taiwan's local factions, see Chao (1997); C. Chen (1997); M. Chen (1995); Huang (2010, 2011); Wang and Huang (2010). 3. Administratively, Taiwan is divided into direct municipalities, counties, and county-level cities. For simplicity, we will often simply refer to both counties and city-level counties as counties. The executives for these are commonly called may- ors and magistrates, respectively. For simplicity, we call them all magistrates. Below the cities and counties are a group of administrative districts including cities, towns, villages, and districts that we will collectively call townships. Township executives 130 Mixed-Member Electoral Systems in Constitutional Context will be called mayors. All these executive posts are elected by the single-member- plurality rule. 4. Similarly, members of the Provincial Assembly and national legislature had access to funds for subsidies to local governments. However, even if the assembly members or legislators could help secure these funds, the money was dispersed by the provincial or central governments instead of the politicians themselves. 5. There are other factions in Taiwan called the Red and Black factions. How- ever, the Taichung County Red and Black factions are not allied with these other Red and Black factions. When Taichung County Red faction members enter the national legislature, for example, they do not necessarily cooperate with Changhua County Red faction legislators or Kaohsiung County Red faction legislators. All three Red factions are independent entities confined to specific counties, and there is no national Red faction. 6. The faction names are drawn from the colors used for campaign materials in the 1950s. Lin He-nian favored red flags, leaflets, and name cards, while Chen Shui-tan used black or dark blue (Hung 1996). These colors may have been chosen to indicate the candidates' social status and perhaps even their attitudes toward tra- ditionalism. In those early days, many of the elites were doctors or teachers. Doc- tors wore Western-style black suits, and were often considered to be more western- ized. Teachers tended to be more conservative and often dressed in red. Chen was a prominent doctor, while Lin was an intellectual from the famous Wufeng Lin clan, which had produced several degree holders (see Meskill 1979). 7. The Red and Black factions do not have any clear policy differences. Individ- ual members may take different positions on particular policies, but these stances do not represent the entire faction. 8. In fact, the KMT actively hastened the decline of the Yang faction. Yang Tien-sheng found himself on the wrong side of the power struggle between Lee Teng-hui and James Soong in the late 1990s, and Yang's business interests suddenly found that government regulators became very hostile. 9. Conventional accounts of the LDP's factions almost never mention local executive races, focusing heavily on legislative races and competition over party leadership and cabinet positions. See Thayer (1969); Curtis (1988); Fukui (1970); Ramseyer and Rosenbluth (1993); and Kobayashi and Tsukiyama (chapter 3, this volume). 10. Even the common story about money does not fit. The usual legislature- centric story involves the legislator building up support by bringing large amounts of public money back to the district. In contrast, Hung and Liu were not particu- larly noted for their ability to extract pork from the provincial or central govern- ments. Rather, both were famous for their ability to penetrate local farmers asso- ciations (Hung 1996). Controlling a farmers association does not necessarily mean that a politician is especially pro-farmer. While the conduit to farming votes is useful, the main advantage to controlling farmers associations is being able to com- mand the financial assets of their affiliated credit unions. However, it is important to remember that local governments offer far more discretionary resources than farmers associations and have always been considered a far bigger prize. 11. For an explanation of this logic, see Batto and Cox, introduction, this volume. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 13 1 12. For an explanation of telephone primaries, see Yu, Yu, and Shoji, chapter 5, this volume. 13. The name of the new direct municipality is actually Taichung City. To avoid confusion with the old, smaller, county-level Taichung City, we call the new entity Taichung Municipality. 14. Appointees must meet the civil service employment qualifications. Most local politicians do not meet these qualifications, so a municipal mayor cannot simply stuff these posts with faction cronies. 15. Taichung County lost 21 township mayor positions, 19 county assembly seats, and 282 township council seats. Taichung County also effectively lost its executive seat, since the new Taichung Municipality mayor was the incumbent from Taic- hung City. In all, 323 elected positions were eliminated. 16. In addition, the party chair, any mayors of direct municipalities, and the three legislative caucus floor leaders are automatically members of the CSC. Mayors of county-level cities and counties are automatically members of the CEC, and they may choose one of their members to sit on the CSC. This system has been used since 2002. Prior to 2002, there were no automatic seats, and there were 31 mem- bers of the CEC who elected 11 members of the CSC. 17. In addition to the four major factions, there were also some smaller and less influential factions in the 1990s. The most famous of these was the Taiwan Inde- pendence Alliance, which had its roots in the overseas independence movement during martial law. 18. Press reports differ, but New Era was generally reported with 15 or 16 legis- lators, while New Dynamic only had about 8 or 9 (see CNA, January 6, 1990; CNA, June 7, 1990) . 19. Not all the pressure came from New Tide. For example, Vice President Annette Lu publicly mooted the possibility of Chen's resignation, and some in the Justice faction urged Chen to hand over power before his term expired. However, New Tide led many of the attacks, and defenders of the president focused their rage on New Tide. 20. The motion ordered factions to dissolve and including five clarifying provi- sions: (1) factions could not maintain offices; (2) factions could not recruit new members; (3) factions could not accept dues from members or raise money; (4) factions could not hold official faction meetings or hear reports from government officials; (5) all party members holding public or party offices should publicly announce they have withdrawn from any factions (LT, July 12, 2006). 21. Money may also have played a role, an interesting possibility given the Japa- nese experience. Raising money has been stressed as one of the main functions of Japanese LDP factions, and one of the main thrusts of 1994 Political Funds Con- trol Act was to prohibit individual factions from directly seeking corporate funds (Reed and Thies 2001; Kobayashi and Tsukiyama, chapter 3, this volume). In the DPP, there was concern that New Tide had been absorbing a disproportionate share of money. Critics pointed to a New Tide leader's position as chair of the Tai- wan Stock Exchange Corporation and the leverage over the financial system that this position bestowed on him. However, there are reasons to doubt that money was really the crucial factor. The other factions also had rainmakers, and many of 132 Mixed-Member Electoral Systems in Constitutional Context them were among the motion's most ardent supporters (CT, July 7, 2006). More- over, New Tide allies proposed an alternate motion that would have prohibited raising money but allowed the factions to continue to exist. The fact that New Tide was willing to sacrifice its fund-raising capacity and that opponents rejected this alternative suggests fund-raising was not the central point. 22. Somewhat comically, immediately after the motion dissolving factions passed, the party congress turned to the last item on the agenda: electing the new CEC and CSC. These elections featured factional politics in all their glory, with factions organizing vote rationing schemes. Of course, the media reported these elections in terms of which factions did well and which ones did not. 23. This does not mean that the dissolution did not affect New Tide at all. Over the final two years of Chen's presidency, observers began to differentiate between the Northern Tide and the Southern Tide. The Northern Tide was far more criti- cal of President Chen and far more vocal about the need for a reform movement within the DPP. Some New Tide members blamed this internal divergence on the fact that they were not able to hold regular meetings and hammer out a common position as they had in the past (CT, May 18, 2007). New Tide only fully reestab- lished an organizational foundation in October 2008 when it founded the New Society for Taiwan think tank. 24. However, two CSC seats were won by factions that are loosely allied with the Chen faction, the Grandparents faction and former Premier Yu Hsi-kun's faction. 25. The DPP has historically not been competitive in several districts, and we considered nominations in these districts to be far less valuable. As such, we do not consider nominations in the following districts: Taitung County, Penghu County, Jinmen County, Lienchiang County, Plains Aborigines, and Mountain Aborigines. Party List and Overseas Representatives are also excluded from this table. REFERENC ES Chao, Yung-mau. 1997. Taiwan Difang Zhengzhi de Bianqian yu Tezhi [The change and characteristics of political ecology on Taiwan's local autonomy]. Taipei: Han Lu Book & Publishing. Chen, Chieh-Hsuan. 1997. "Faction Network, Electoral Cadre Network, and Civil Network: The Social Meaning of a Local Faction's Initiation" [in Chinese]. In Difang Shehui [Local Society], ed. Institute of East Asian Societies and Econo- mies, Tunghai University. Taipei: Linking Book Press. Chen, Ming-Tong. 1995. Paixi Zhengzhi yu Taiwan Zhengzhi Bianqian [Faction poli- tics and political change in Taiwan]. Taipei: Yue-dan. Cheng, Ming-teh. 2004. Yi Mai Zong Xiangcheng: Paixi Zhengzhi zai Minjindang [Faction politics in the DPP]. Taipei: Shiying. Cheng, Ming-teh. 2010. "Why Did the DPP National Party Congress Dissolve the Factions?" [in Chinese]. Meilidao Dianzi Bao [Formosa] April 30. http://www. my-formosa.com/article.aspx?cid=5&id=4006 (accessed March 26, 2012). Cheng, Tun-jen, and Yung-ming Hsu. 1996. "Issue Structure, the DPP's Factional- ism, and Party Realignment." In Taiwan's Electoral Politics and Democratic Transi- tion: Riding the Third Wave, ed. Hung-mao Tien. Armonk, NY: M. E. Sharpe. Executive Competition, Electoral Rules, and Faction Systems in Taiwan 13 3 Curtis, Gerald L. 1988. TheJapanese Way of Politics. New York: Columbia University Press. Fukui, Haruhiro. 1970. Party in Power: The Japanese Liberal-Democrats and Policy- Making. Berkeley: University of California Press. Huang, Hsin-ta. 2010. "An Analysis of the Possible Effects on the Local Political Ecology of the New Taichung Municipality" [in Chinese]. In Gonggong Zhengce yu Difang Zhili: Difang Zizhi Lunwenji [Public policy and local governance: Autonomy in local government], ed. Hsing-chou Sung and Chien-jen Chen. Taipei: AIRITI. Huang, Hsin-ta. 2011. "A Study of the Relationships between Local Factions and Political Parties after the Taichung County-City Consolidation" [in Chinese]. Zhengzhiyu Zhengce [Asian Journal of Politics and Policy] 1 (2): 91-140. Huang, Hsin-Ta, Yeh-Lih Wang, and Chien-Hsiung Wang. 2009. "The Impact of Electoral Reform on Local Politics: A Case Study of Taichung County." Paper presented at the International Conference on Elections in Taiwan, Japan, and Korea under the Mixed-Member Electoral System, NCCU University, Taipei City, Taiwan, May 24. Hung, Chun-mu. 1996. "Taichung County: The Continuing Lin-Chen Rivalry and the Rise of the Third Force" [in Chinese]. In Difang Paixi yu Taiwan Zhengzhi [Local factions and Taiwan politics], ed. Kun-shan Chang and Cheng-hsiung Huang. Taipei: Lianjing. Jacobs, J. Bruce. 1979. "A Preliminary Model of Particularistic Ties in Chinese Political Alliances: Kan-ch'ing and Kuan-hsi in a Rural Taiwanese Township." China Quarterly 7 8:2 37-7 3. Jen, Yu-te. 2006. "The Establishment and Development of KMT's Electoral Machine and Mobilization System during the 1950s" [in Chinese]. Zhengzhi Daxui Lishixue Bao [Journal of History] 25:71-116. Kao, Yuang-Kuang. 2000. "Disparities of Urban-Suburban Development and Political Factions: A Case Study to Local Elections of Taipei County in 1998" [in Chinese]. Xuanju Yanjiu [Journal of Electoral Studies] 7 (1): 53-86. Kao, Yuang-Kuang. 2004. "An Analysis to the Interactive Pattern of Political Fac- tions and Gangsters in Taipei County" [in Chinese]. Xuanju Yanjiu [Journal of Electoral Studies] 11 (1): 33-72. Kohno, Masaru. 1992. "Rational Foundations for the Organization of the Liberal Democratic Party in Japan." World Politics 44:369-97. Meskill, Johanna M. 1979. A Chinese Pioneer Family: The Lins of Wufeng, Taiwan, 1729-1895. Princeton: Princeton University Press. Ramseyer, J. Mark, and Frances McCall Rosenbluth. 1993.,Japan's Political Market- place. Cambridge: Harvard University Press. Reed, Steven R., and Michael E Thies. 2001. "The Consequences of Electoral Reform in Japan." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg. New York: Oxford University Press. Rigger, Shelley. 2001. From Opposition to Power: Taiwan's Democratic Progressive Party. Boulder: Lynne Rienner. Scott, James, C. 1972. "Patron-Client Politics and Political Change in Southeast Asia." American Political Science Review 66 (1): 91-113. 134 Mixed-Member Electoral Systems in Constitutional Context Thayer, Nathaniel B. 1969. How the Conservatives Rule Japan. Princeton: Princeton University Press. Wang, Chin-shou. 2004. "Returning to Feng-mang County: The Success and Fail- ure of the Kuomintang Political Machine" [in Chinese]. Taiwan Zhengzhixue Kan [Taiwan Political Science Review] 8 (1): 99-146. Wang, Yeh-lih, and Hsin-ta Huang. 2010. "Local Factions after Twin Transitions of Government in Taiwan: Decaying or Transforming?" Paper presented at the Annual Meeting of the American Political Science Association in Washington, DC, September 2-5. Wu, Chung-li. 2002. "Controversy about the Research on Taiwan's Faction Poli- tics: Learning from the Analytical Approaches of American Machine Politics" [in Chinese]. Zhengzhixue Luncong [Political Science Review] 17:81-106. Wu, Nai-teh. 1987. "The Politics of a Regime Patronage System: Mobilization and Control within an Authoritarian Regime." PhD diss., Department of Political Science, University of Chicago. Newspaper articles are cited by source and date: CNA: Zhongyang She (Central News Agency) CT: Zhongguo Shibao (China Times) LT: Ziyou Shibao (Liberty Times) UDN: Lianhe Bao (United Daily News) FIVE Innovations in Candidate Selection Methods Eric Chen-hua Yu, Kaori Shoji, and Nathan F Batto This chapter focuses on how major parties adjusted their candidate selec- tion methods (CSMs) to meet the challenges brought about by the new electoral regimes in Taiwan and Japan, respectively. Specifically, it inves- tigates two innovative ways to select district-level candidates adopted by the major parties in each country.' In Taiwan, the polling primary became the default system that the two major parties, the Kuomintang (KMT) and the Democratic Progressive Party (DPP), used to nominate their candi- dates for the Legislative Yuan elections. In Japan, both of the major par- ties, the Democratic Party of Japan (DPJ) and the Liberal Democratic Party (LDP), have turned to the kobo (public recruitment) system to select candidates. The new CSMs were designed to address different concerns. The polling primary was intended to identify the strongest candidate and resolve intraparty competition. In contrast, the initial purpose of the kobo system was not to resolve intraparty clashes among ambitious and powerful contenders. Rather, in the late 1990s and early 2000s the DPJ used kobo for the purposes of party-building. In more recent years, both the DPJ and LDP have increasingly used it to legitimize party decisions. Candidate Selection Methods, Electoral Reform, and Constitutional Systems The determinants of CSMs is an important topic for scholars interested in political parties (Ranney 1981; Gallagher and Marsh 1988). Recent work 135 136 Mixed-Member Electoral Systems in Constitutional Context on the extent to which electoral systems affect the choice of CSMs has not pointed to a consensus. While some comparative case studies argue that electoral institutions should have substantial impacts on choices of CSMs, particularly in terms of ballot structure and district magnitude (Norris 1997; Kasapovic 2001), other recent large-N studies show that the empirical evidence is meager at best (Lundell 2004; Shomer 2012). From the perspective of level of analysis, Hazan and Voerman (2006) point out that electoral systems are a country-level variable that may not account for intracountry, cross-party variation in CSMs. Gallagher and Marsh (1988) are perhaps correct in positing that electoral systems do not, by themselves, completely determine CSMs, yet it is certainly possible that electoral systems exert some degree of influence. Decades of research on electoral systems has clearly indicated that parties do, in fact, vary their strategies according to the different incentive schemes produced by dif- ferent electoral systems. CSMs are an important element of overall party electoral strategies, and we should expect to see some impact of the elec- toral incentives on the choice of CSMs. In the particular cases of Taiwan and Japan, although both the polling primary and kobo had been initiated prior to the introduction of electoral reforms, we argue that the electoral and constitutional systems helped facilitate the consolidation of both inno- vative institutions. Changing from SNTV to MMM may push parties to adjust their CSMs. For one thing, such a change increases the importance of managing intraparty competition in the nomination stage (Wu and Fell 2003; Krauss and Pekkanen 2004). Under SNTV, much of the intraparty competition takes place in the general election stage, as parties routinely nominate mul- tiple candidates in the same district. While the optimal strategy is for a party to nominate the "right" number of candidates and then distribute its support evenly among them (Cox and Rosenbluth 1993; Cox and Thies 1998; Cox and Niou 1994; Browne and Patterson 1999), overnomination is not always a disaster. If more than the appropriate number of candidates insists on running and no compromise can be found, the party can some- times simply let them all run and let the electorate decide which is the weakest. That is, the party can opt for a less optimal solution in which no candidate is unfairly denied a nomination. Sometimes this will lead to a worse outcome, but sometimes the party will still manage to avoid losing a winnable seat. In contrast, in single-member districts (SMDs), parties must nominate a single candidate to have any chance to win. A second candi- date in an SMD race almost always portends disaster. Thus, any intraparty clashes that appear during the nomination process have to be dealt with at that stage and cannot be pushed off until the general election. Innovations in Candidate Selection Methods 137 Another way in which the shift to MA/M could affect nominations involves the increased number of districts. Under SNTV, each district elects multiple seats. Any party wishing to give every voter a chance to support it has merely to nominate one candidate in every district. With the change to MMM, the number of districts increases markedly, and parties wishing to give all voters an option to support them have to nominate a much larger number of candidates. Even for established parties, this is not an easy task. For smaller and newer parties without deep pools of potential candidates, simply finding enough competent candidates to run in every district can be a daunting challenge. Especially in weaker districts, newer and smaller par- ties have to struggle to present a roster of credible candidates. Japan's parliamentary system creates a different set of problems than Taiwan's semipresidential system does. Batto and Cox (introduction, this volume) and Lin (chapter 2, this volume) argue that executive posts are the ultimate goal for many politicians. In Japan, cabinet posts are deter- mined by negotiations among party leaders in the House of Representa- tives. Almost all of the key figures, including the party leaders and the MPs who receive ministerial posts, are legislators who have accumulated years and years of seniority. The typical career path is well documented. An MP is elected at a young age and slowly works his way up through a series of posts. After being reelected several times, he or she might finally accumu- late sufficient seniority and experience to be eligible for a ministerial post. Party leaders are generally drawn from this same group of senior MPs. In short, seniority is a prerequisite for securing powerful posts. Taiwan's semi- presidential system works differently. Cabinet posts are allotted unilater- ally by the president, not negotiated by legislative leaders. Moreover, since a legislator must resign his or her seat to assume a post in the executive branch, presidents are reluctant to appoint legislators to the cabinet. Ambi- tious legislators must thus look elsewhere if they hope to hold executive power. In fact, ambitious legislators typically seek to win local executive seats as city mayors or county magistrates. National leaders generally have proven their mettle by winning a highly competitive mayoral race and suc- cessfully administering a city government for several years. The legislators who remain in the legislature and accumulate seniority are generally less energetic, charismatic, and politically talented than those who move on to other challenges. The result of this is that legislative seniority is much more important in Japan than in Taiwan. Most important Japanese politi- cians are senior legislators, while few Taiwanese leaders are. The differing importance of seniority has significant implications for nomination strategies. Consider a district in which a party is very strong. The nomination in this party is a very valuable commodity, since nomina- 138 Mixed-Member Electoral Systems in Constitutional Context tion almost certainly means winning in the general election. Such a valu- able commodity is certainly worth fighting over, even if it means taking on an entrenched incumbent. In Japan, all of the senior figures in the party have accumulated seniority, and they all have an interest in protecting that precious asset. To this end, Japanese parties have written the rules to favor incumbents. In particular, Japanese parties typically automatically renomi- nate incumbents. In Taiwan, the decision makers in the party do not usually have strong personal interests in protecting seniority. They would gener- ally prefer to see the strongest candidate emerge. In fact, party leaders may actually prefer to see energetic challenges to ensure that incumbent legisla- tors work hard and actively stay in touch with their constituents. As a result, Taiwanese and Japanese CSMs have evolved to address very different challenges. For Taiwanese parties, the most important problem is how to choose nominees in the strongest and most desirable districts. The polling primary has emerged as the best solution for this task. As par- ties cannot always easily resolve intraparty clashes, they have an incentive to delegate the power to the general public to make the decision. Public opinion surveys are seen as an objective test of strength in which everyone gets a fair shot. Surveys have the added advantage of communicating a quantifiable measure of strength and weakness to the general public. Voters who support the loser not only have to face evidence that their favored can- didate has lost in a fair process but also that if he ran a renegade campaign in the general election he would have to face the prospect of widespread strategic voting toward the party nominee. Japanese parties have a different problem. The most desirable districts are almost all occupied by incumbents, and intraparty challenges for these nominations are simply disallowed. Instead, Japanese parties worry about nominating candidates for districts without incumbents. Many of these dis- tricts are hopeless and undesirable to ambitious politicians. Nevertheless, big parties generally wish to run candidates in all or almost all districts, so they try to find someone to represent the party.2 In the years just after the electoral reform, this was a particular challenge for the DPJ. The old pri- mary opposition, the Japan Socialist Party (JSP), disintegrated in the new electoral system as it was unable to win pluralities in many districts. The DPJ rose to fill this void, but, as a new party, it did not have deep pools of candidates ready to run in every district. The DPJ turned to kobo to recruit new politicians and ensure that it could run a full slate of candidates. At first, the DPJ used kobo primarily as a party-building mechanism. As the DPJ matured and finding enough candidates to run nationwide became less of a problem, the nature of kobo began to shift. Many districts with no Innovations in Candidate Selection Methods 139 incumbents were worth fighting for. Local power holders or national fac- tion leaders might maneuver to nominate their most favored candidate into these spots, but concerns over fairness spurred them to try to cloak their influence. Increasingly kobo was used to publicly legitimize decisions made by party leaders. That is, kobo was presented to the public as an open and fair decision-making process. In practice, it was anything but open. Even the LDP, a well-established party with an entrenched local party organi- zation, shared some similarities to the DPJ in the incentives for adopting kobo. Below, we look into the development of new CSMs in each country. The Polling Primary in Taiwan The Evolution of Taiwanese CSMs and the Emergence of the Polling Primary The KMT implemented an authoritarian regime after retreating to Tai- wan in 1949 and institutionalized different types of elections in order to facilitate its control at the grassroots (Cheng 1989; Wu 2001). To solidify its authority in elections, the KMT adopted a top-down CSM. During the authoritarian era, local party members were informally consulted in the candidate selection process, but the overall CSM was characterized by centralization. During the transition to democracy in the mid-1980s, Taiwan saw a series of important political markers, such as the establishment of the main opposition party, the DPP, in 1986, the dismantling of martial law in 1987, and the death of President Chiang Ching-kuo in 1988. These events sig- naled the decline of the KMT's authoritarian rule and the emergence of competitive party politics (Wu 2001). In response to the inevitable trend of democratization, the KMT adopted and started transforming its Leninist party structure to compete in the new democratic structure. One of the starting points was to open and decentralize its CSM. In 1989, the KMT adopted a closed primary system. This was the first time that the party's rank-and-file members could formally participate in the candidate selection process and marked a major decentralization of author- ity. While the central party headquarters continued to retain the right to make the final decisions, it generally respected the primary results. How- ever, this new system did not produce an entirely favorable outcome. One of the KMT's goals was to reduce the power of local factions, and this new system led to more nominees who were not associated with local factions 140 Mixed-Member Electoral Systems in Constitutional Context (Huang 1996). The KMT share of elected seats fell by 9.5% from 1986, and the new CSM, which many argued had produced a weak slate of candidates, was commonly cited as one of the reasons for this poor performance. The KMT revised its CSM for the 1992 elections to include party member primaries followed by evaluations from party cadres. While this new system was somewhat less decentralized than the 1989 system, the KMT did not return to a fully centralized CSM. Rather, the 1992 system confirmed the power shift from the center to localities, and local party members and party officials have exercised significant influence in deter- mining nominations in all subsequent elections. In the 1993 county magistrates and city mayoral elections, the KMT further adjusted its CSM to include three factors: party member opinion responses, evaluations by cadres, and public opinion polls (Wu 2001). This marked the first time that the KMT incorporated polling results in its can- didate selection process. However, the results from these three processes were not binding, as higher party officials reserved the right to make the final nomination decisions. Although the KMT was the first party to use polls to assess aspirants, the DPP was the first party to incorporate polling results into a binding formula to determine nominations. In the early years after its founding in 1986, the DPP's leadership was not composed as a unified group but came from various anti-KMT individuals or factions that occupied offices at the local representative bodies. This unique preparty history gave the DPP a decentralized pattern of power distribution that was reflected in its CSM. Negotiations and compromises among party factions became the major tool for nominating candidates. If party leaders were unable to reach any consensus, party members were to be asked to cast votes to determine final nominations. In practice, the leadership of the DPP tried to reach consen- sus through compromise, and it was rare to see party members voting in the 1980s. The above two-stage process characterized the DPP's candidate selection from its founding until the early 1990s, though it was not offi- cially codified until the 1992 legislative election. In 1995, the DPP revised its CSM to give party cadres more power.3 If the party failed to settle on a consensus, votes of party members and party cadres each accounted for 50% of the final decision. The introduction of voting for party cadres signaled the increasing importance of local party branches and a deemphasis on party members in the nomination process. One reason for this was to address an increasing tendency for aspirants to register large numbers of new party members. The aspirant would typi- cally pay the party dues and control the votes of these "head voters" (rentou Innovations in Candidate Selection Methods 141 dangyuan), and many DPP leaders and supporters saw this as a growing form of corruption or vote buying (Wang 2006). Others worried that DPP factions were becoming too powerful, and it was increasingly impossible for aspirants not affiliated with a faction to win a nomination. In fact, the addition of party cadres did little to reduce the influence of factions, as most of the cadres were themselves deeply embedded in the faction system. The DPP strategy for avoiding these negative repercussions was to decentralize their nominations. Instead of relying on the small number of formal party members or an even smaller number of party leaders to make decisions, the DPP attempted to empower the much larger number of party supporters and sympathizers in the general public. Aspirants could manipu- late the outcomes inside the party by registering easily controlled "head voters" or by factional maneuvering, but the hope was that these strategies would not be feasible in the much larger population of party supporters. This new strategy was first attempted in the contest for the 1996 presiden- tial nomination. The DPP used a closed primary to winnow the field to the top two contestants, and then these two toured the country in a series of debates in which the general public was allowed to vote to determine the nominee. Unfortunately, the attendance at these open primary events was still fairly low, so the DPP continued to look for a better process. This led to the incorporation of telephone polling in the 1998 legisla- tive nomination process. The DPP scrapped voting by party cadres and instead gave 50% weights to both party member votes and telephone poll results. This marked the first time that either of the two major parties had included polling results in the nomination process in a binding, not merely an advisory, role. During the 1980s and 1990s, the two main parties' CSMs evolved in very different ways. The KMT's main concern was to devolve power from the center to localities in order to compete more effectively in the new democratic system, while the DPP moved from an elite-dominated CSM to one that drew on a much larger selectorate in an attempt to mitigate manipulation of its internal party processes. However, once the polling primary was introduced and its advantages became apparent, the 2000s saw the two parties converge on relatively similar CSMs. In the 2001 Legislative Yuan election, the KMT imitated the DPP by adopting polling results in its nomination process. The KMT's CSM included two parts: voting by local party members and telephone polling, each given equal weights. Local party branches tabulated and combined the results of the two parts and forwarded them to the central party head- quarters, which generally respected the local results. TABLE 5.1. Evolution of CSMs for Legislative Yuan Elections, 1969 to 2012 Year KMT DPP 1969 Candidates produced by consultations with local party members, then reviewed by the provincial party headquarter, and finally decided by the central party headquarters 1972 Candidates produced by consultations with local party members, then decided by the central party headquarters 1975 Same as 1972 1980 Candidates produced by consultations with local party members and reviews of local party officials, then reviewed by the provincial party headquarter, and finally decided by the central party headquarters 1983 Candidates produced by consultations with local party members and reviews of local party officials, then the provincial party headquarter doubled the number of nominee, and finally decided by the central party headquarters 1986 Same as 1983 1989 Candidates produced by party member primary, then decided by the central party headquarters 1992 Candidates produced by party member primary (50%), and reviews of local party officials (50%), then decided by the central party headquarters 1995 District candidates produced by consultation with local party members, or by opinion polls, or by reviews of local party officials, then decided by the central party headquarters 1998 Same as 1995 2001 Candidates produced by party member primary (50%), and opinion polls (50%) 2004 Candidates produced by party member primary (30%), and opinion polls (70%) 2008 Same as 2004 2012 Candidates produced by opinion polls N/A N/A N/A N/A N/A N/A Candidates produced by consensus; otherwise by party member vote Candidates produced by consensus; otherwise by party member vote Candidates produced by consensus; otherwise by party member vote (50%) and party cadres vote (50%) Candidates produced by compromises (consensus), otherwise by party member vote (50%) and opinion polls (50%) Candidates produced by compromises (consensus), otherwise by party member vote (30%) and opinion polls (70%) Same as 2001 Same as 2001 Candidates produced by opinion polls Source: Data before 2012 were adopted and revised from Wang (2008, 143-70). Note: The nomination of party list candidates is not included. All nomination decisions, even those after parties began using polling primaries, had to be ratified by the national party headquarters. Innovations in Candidate Selection Methods 143 The importance of polling results continued to increase. The DPP adjusted its formula to make telephone polling account for 70% of the final outcome, with party members voting reduced to only 30%. The KMT fol- lowed suit in 2004, and both parties used the 70-30 rule to determine their district-level nominees for the 2004 and 2008 Legislative Yuan elections.4 In 2012, both parties completed the evolution toward the polling primary by eliminating voting by party members and determining nominations solely by the results of telephone polls.5 While the polling primary evolved in the SNTV era, it arguably is even better suited to nominations for the single-seat districts employed in MMM. In a large SNTV district, a party wishing to nominate four candi- dates might have to adjudicate between its fourth and fifth strongest aspi- rants. These aspirants might not be strong enough to cause problems if not nominated, and the party always has the option to simply overnomi- nate. In MMM, the clash is between the two strongest aspirants and over- nomination is not a feasible option. Resolving the conflict in a fair way is imperative. Moreover, SNTV elections are characterized by intraparty competition, so finding a balanced roster of nominees that appeal to vari- ous factions within the party is important. Internal party processes, such as a closed primary, are useful for ensuring that no single faction monopolizes nominations. In contrast, internal party balance is less urgent in MMM. Single-seat elections are characterized more by interparty competition, and the strongest candidates are those that appeal to broad swaths of the population. Polls are particularly adept at identifying these politicians. In fact, while this chapter focuses on legislative nominations, both parties have used polling primaries extensively for single-seat executive elections. It is thus not surprising that the parties responded to electoral reform by intensifying their use of the polling primary. The Mechanics of the Polling Primary This section describes the CSMs used by KMT and DPP in the 2012 leg- islative election. Both the KMT and DPP started their candidate selection procedure with candidate registration. All aspirants had to register, includ- ing incumbents. In general, incumbents were not given any special sta- tus in the competition for nominations, and sitting legislators often faced competition from challengers within the party. Even when only one candi- date registered, he or she still needed to demonstrate electability through public opinion polls. The KMT required that unopposed aspirants must pass a 30% threshold of support to win a nomination, while the DPP head- 144 Mixed-Member Electoral Systems in Constitutional Context quarters reserved the right to reject weak aspirants and search for stronger alternatives. If more than two aspirants registered for nomination selection, the pro- cess proceeded into the second phase-negotiation. Both parties preferred to resolve conflicts behind closed doors rather than by engaging in open and potentially explosive intraparty competition. During the negotiation period, parties sometimes did nonbinding polling to see who had a bet- ter chance to win and who should be encouraged to yield. Additionally, the party headquarters sometimes sent a senior party official to handle the negotiation process. In some cases, such negotiation processes was suc- cessful and the party was able to nominate a candidate without intense and direct intraparty conflict. For example, in 2012 in New Taipei City 4th District, the KMT's incumbent Lee Hung-chun faced a challenge from Hsu Bing-kuan, who had a strong local organization and significant grass- roots support. Lee and Hsu had known each other for 30 years and were close friends. In the beginning of the negotiation process, Lee expressed a willingness to withdraw and let Hsu represent the KMT in the general election. Yet, the KMT leadership favored Lee due to both his perfor- mance in the legislature and his broad popularity in the very competitive 4th District. Thus, at the end of the day, Hsu agreed to withdraw and wait on the sideline for future opportunities.6 Most negotiations did not go as smoothly as the previous example. Quite often, contenders refused to step aside and the negotiation pro- cess broke down. For example, in New Taipei City 2nd District, the DPP incumbent Lin Shu-fen was challenged by Huang Chien-hue, a former legislator who had lost to Lin in the 2008 DPP primary. Soon after the deadline for registration, the DPP headquarters sent a senior party leader to initiate the negotiation process. Yet, the negotiations fell apart quickly as Huang, who was strongly supported by one of the DPP's major fac- tions, made it clear that he would never withdraw from the competition.7 In short, although Lin seemed to be a very strong DPP incumbent candi- date, a polling primary was still necessary since a quality contender insisted on challenging her. When negotiations broke down, the process moved into one of two third phases. In stronger districts, parties held a polling primary, while in weaker districts the party headquarters tended to directly select a nominee. In weaker districts, the party had little chance of winning and the challenge was more commonly to simply recruit any quality candidate rather than to mediate between multiple competent aspirants. In these more difficult districts, the party headquarters sometimes unilaterally decided who the Innovations in Candidate Selection Methods 145 nominee would be. For example, the DPP deemed all districts in which it had received less than 42.5% of the vote in 2008 as "difficult," and the party headquarters was given the authority to unilaterally decide nominations in these districts. Where parties were stronger, the nominations were more valuable, intraparty conflict was potentially more explosive, and parties were more likely to end up with the polling primary. Table 5.2 shows that in nine of the 11 districts in which the KMT held a polling primary in 2012, it even- tually won the seat. The DPP was only slightly less successful when it held polling primaries, winning nine of 13 districts. Incumbents were not immune from challenges. Twelve incumbents were forced to face a polling primary, and three lost. Fifty KMT and DPP incum- bents were eventually nominated without resorting to a polling primary. However, this should not be interpreted as 50 unchallenged nominations. In many cases, other aspirants registered or considered registering but were eventually persuaded to yield before or during the negotiation phase. There were many similarities and a few clear differences in the protocols TABLE 5.2. Use of Polling Primaries in the 2012 Legislative Yuan Election KMT DPP Number of seats % of total Number of seats % of total Total single-seat districtsa 73 100.0% 73 100.0% Districts nominated 71 97.3 71 97.3 Districts won in 2012 44 60.3 27 37.0 election Districts held prior to the 50 68.5 19 26.0 2012 electionb Districts using a polling 11 15.1 13 17.8 primary Polling primary winners 9 12.3 9 12.3 who also won the seat Incumbents nominated 39 53.4 11 15.1 without a polling primarye Incumbents participating 7 9.6 5 6.8 in a polling primary Incumbents winning a polling primary 5 6.8 4 5.5 aOnly single-seat districts are considered in this table. bThis is slightly different from the 2008 election result because there were several by-elections between the two general elections. 'For this table, an incumbent is defined as a legislator who won the district in the 2008 election or an intervening by-election. Legislators who won their seat in a different district or on the party list are not considered incumbents. 146 Mixed-Member Electoral Systems in Constitutional Context used by the two parties (table 5.3). Both parties contracted with outside polling companies to do random sample surveys. The DPP had a list of eligible polling organizations, including its own internal survey unit, and randomly selected three of these for each different race. The KMT adopted a more decentralized method, allowing any two polling organizations to be selected in any race upon the aspirants' agreement. The two parties applied different scheduling of the surveys, and the KMT employed in-house sam- pling8 while the DPP did not. These two differences reflected the different priorities of the two parties. The KMT placed the highest priority on get- ting an accurate reading of public opinion, and its protocols were designed to ensure a representative sample. The KMT prescheduled its surveys to be held over three days, usually including both weekdays and weekends, and it further used in-house sampling rather than simply asking whoever answered the phone in order to eliminate as much as possible any distor- tions in the sample. The DPP, in contrast, saw the polling primary as a way to test the mobilization capacity of its contenders.9 Rather than try- ing to eliminate any distortions, the DPP encouraged its contestants to try everything possible to affect the outcome. The scheduling, in particular, was designed specifically to test mobilization. The headquarters only presched- uled a period of time for a list of district primaries without specifying the order. Thus, no one knew which primary would take place on what particu- lar date. Then every morning during the primary period, the headquarters randomly selected one district from the list for a polling primary that night. The contenders were immediately informed, usually around 10:00 a.m. Once the primary date was announced, contenders rushed to mobilize their supporters, encouraging them to be in their homes between 6:00 and 10:00 p.m. that night to wait for possible phone calls. The DPP did not adopt in- house sampling for the same reason. In households with split preferences, TABLE 5.3. Key Features of KMT and DPP Polling Primaries in 2012 KMT DPP Overseen by: Local party branch National party headquarters Method: Telephone survey Telephone survey Days to complete survey: 3 1 Survey organizations: 2 3 Interviews completed by each At least 1,068 About 1,200 organization: Questionnaire: Interparty matchup: 85% Default: Interparty matchup Intraparty matchup: 15% (can be changed by consensus) Date: Prescheduled Random In-house sampling: Yes No Innovations in Candidate Selection Methods 147 who answered the phone became an important factor in determining the results. Contenders had just a few hours to phone, text, or otherwise contact as many people as possible to ensure that a disproportionate number of their supporters were at home ready to answer the phone. Finally, the questions put to respondents varied across parties. The DPP's default question in strong districts was an interparty comparison.'0 For example, if there were two DPP contestants (A and B), respondents were typically asked a pair of questions, whether the respondent preferred A or the KMT candidate and whether the respondent preferred B or the KMT candidate. However, if the contestants agreed on some alternate for- mat or if the KMT candidate had not yet been determined," the DPP was willing to allow other question formats. Unlike the scheduling, which the party headquarters kept under tight control, the question format was open to revision. One common complaint about the interparty comparison was that it did not allow sincere party loyalists to express their preferences for one or the other contestant. Since many loyalists prefer any DPP candi- date to any KMT candidate, sincere respondents would express support for both A and B and thus have no impact on the outcome. Only those who strategically answered that they would support the KMT candidate over the less favored DPP contestant affected the outcome, and this led to com- plaints that the system forced respondents to lie if they wished to have any impact. The KMT system addressed this concern by including both inter- party and intraparty comparisons. In addition to pitting the contestants against the DPP nominee,'2 respondents were further asked which of the KMT contestants they preferred. The interparty comparison was still the more important, accounting for 85% of the final score, but the 15% weight given to the intraparty comparison allowed KMT supporters to show their preference for a specific KMT contestant. Table 5.4 shows an example of a polling primary. In 2010, the DPP had won the seat in Taitung County in a by-election, and the incumbent, Lie Kuen-cheng, wanted to run for reelection. He was opposed by the for- mer deputy head of the county government, Liu Chao-hao. Negotiations did not yield any compromises, so a polling primary was conducted. Since the KMT had already settled on a candidate, Yao Ching-ling, the standard interparty questions were used. On the morning of April 6, 2011, the DPP headquarters randomly drew the Taitung County district for that day's polling primary. At the same time, it randomly drew three survey organiza- tions to carry out the polls. Between 6:00 and 10:00 that night, each orga- nization conducted 1,200 interviews. The DPP issued a press release on the morning of April 7 announcing that Liu had defeated Lie and would be 148 Mixed-Member Electoral Systems in Constitutional Context nominated. Liu's nomination was confirmed later that week at the weekly meeting of the DPP's Central Standing Committee. Two points about the results are worth noting. First, the results from the three organizations were very similar. This was almost always the case, and similar results from different organizations helped to assuage any suspicions of manipulation. Second, no consideration was given to ideas about margin of error. Results were reported with an unrealistic degree of precision, and there was no requirement that the margin of victory be over a certain threshold. Discussion There are several reasons that the two main parties have turned to the poll- ing primary as the default CSM. First, the polling primary controls costs for parties. Commissioning several polls costs money, but it is cheaper than organizing party member primaries.'3 Second, polling primaries identify the candidate who is the most popular among the general public and thus most likely to win in the general election. Previous CSMs relied on more narrow segments of the electorate to decide nominations, and this left them open to the charge that they were not selecting the best candidates. For example, in the early 1990s the KMT leadership "parachuted" several can- didates with no local ties into districts, and these were often attacked as out- siders and easily defeated. With the polling primary, parachute candidates whose national fame did not translate into local popularity were rejected while those who did enjoy substantial local support were legitimized. Third and most important, the polling primary is seen as a much fairer system than any other alternative. Polls produce hard numbers and a clear decision rule: candidates with support from more respondents win. The fact that the poll results are usually very similar across multiple survey organizations gives them added credibility. Moreover, poll results are much TABLE 5.4. Polling Primary for DPP's 2012 Legislative Nomination in Taitung County All Dimension DPP Survey Master Survey & Survey & Survey organization Center Research Research Average Candidate DPP Yao DPP Yao DPP Yao DPP Yao Liu .4175 .2368 .4195 .2481 .4487 .2332 .4286 .2394 Lie (incumbent) .3823 .2799 .3905 .2913 .4198 .2876 .3975 .2863 Source: DPP website (http://www.dpp.org.tw/newscontent.php?&sn=4844), accessed July 2, 2013. Notes: The surveys were conducted on April 6, 2011. Results were announced on April 7. Respondents were asked a pair of questions, whether they preferred Liu or Yao (the KMT candidate) and whether they preferred Lie or Yao. Innovations in Candidate Selection Methods 149 harder to manipulate than closed primaries or negotiations by party elites behind closed doors. The KMT was particularly worried about the influence of local factions in its nomination process. Factions were extremely effective in closed-room negotiations and could often control the recommendations forwarded to the national party by local party branches. This ability to control the pro- cess was seen as somewhat illegitimate by the general public, as the factions appeared to simply be dividing up spoils among themselves, and opportu- nities for regular people to influence nominations were limited. By moving the locus of competition from party elites to the general public and forcing contestants, including faction members, to compete in a transparent pro- cess, the KMT was able to credibly claim that faction-affiliated contestants had no unfair advantages. The DPP was also concerned about the influence of its factions, which were also seen as somewhat illegitimate by many party supporters. In the DPP, the question of faction power was closely related to the number of phantom party members. Most powerful DPP figures controlled a certain number of "head voters," and they relied on these votes both in internal party elections and in nomination fights. Indeed, DPP factions all had large numbers of head voters in their pockets. Since the formal membership of the DPP was relatively small, factions or individual politicians could decide the outcome of closed primaries by recruiting more head voters, and this was widely seen as a perversion of democracy. The polling primary ren- dered the "head voter strategy" obsolete. More generally, one can see these phantom party members as a type of vote-buying strategy, and, by dramati- cally increasing the size of the selectorate, the polling primary made vote- buying extremely inefficient. In fact, given the low payoff, the negative consequences resulting from being publicly exposed as a vote-buyer, and the high probability that one's opponents would uncover and expose any such activity, the polling primary arguably transformed vote-buying from a beneficial strategy into a harmful one.'4" There are some criticisms of the polling primary (Wu 2008, 115-45). One popular normative argument is that parties have a responsibility to nominate good people, and by adopting the polling primary and delegating decisions to the general public, parties are abdicating this duty.16 A related argument is that allowing the general public to make the decision leaves no important role for ordinary party members. Why should a person for- mally join a party if he or she has no say in its most important decisions? According to this argument, the polling primary could negatively affect party-building. 150 Mixed-Member Electoral Systems in Constitutional Context Another criticism is that the polling primary encourages politicians to be overly concerned with public opinion. Some Taiwanese commen- tators have suggested that the polling primary makes potential aspirants (including incumbent candidates) care more about their public image than anything else. As such, politicians expend too much effort seeking media exposure and building name recognition and not enough energy working on mundane but important policy questions. A quite different line of criticism is more technical, arguing that survey results may not accurately reflect popular opinion. On the one hand, it takes quite a bit of expertise to conduct a random sample survey. Surveys include many different components, such as sampling technique, question- naire wording and ordering, data processing, and so on. Each component is subject to manipulation, and slight variations may produce different out- comes. In the worst case scenario, an unrepresentative sample or an inten- tional distorted outcome might give a victory to the wrong person. Less dramatically, surveys produce estimates with statistical errors. Any victory, no matter how small, is sufficient to win the nomination. In close races, the margin of victory may be within the statistical error, but the parties do not consider this. On the whole, however, the polling primary is generally viewed as a positive innovation. Even if it is not perfect, it has produced substantial benefits to the DPP and KMT by institutionalizing a system that resolves intraparty conflict in a fair and transparent manner. The Kobo System in Japan The Evolution of Japanese CSMs and the Emergence of the Kobo System The Japanese term kobo literally means public advertisement of a post. The opening of a position is publicly announced, and anyone who is interested and qualified can apply. The term only suggests that the entry to the selec- tion process is open to public, but other important aspects could vary (and are often left unspecified), such as what level in the party organization will make the decision, how much outside participation will be allowed, and what the criteria for the decision will be. This method, which merely declared that anyone could enter the nomination race, was regarded as an innovative CSM only in comparison to the closed and informal traditional Japanese nomination practice. Innovations in Candidate Selection Methods 151 Prior to the appearance of kobo in the 1990s, there was no formal struc- ture for candidate selection adopted by major political parties in Japan. Each party would recruit its candidates through its internal networks, the opening was never officially announced, and the entry to the selection was never made open to the public. For example, prior to the 1994 electoral reform, the LDP typically delegated recruiting candidates to each of its factions (chapter 3, this volume; Krauss and Pekkanen 2011). The supply of human resources came mainly from local and national electable office- holders, national-level bureaucrats, and the heirs of MPs. For other smaller parties, it was the norm to recruit candidates from affiliated organizations, such as trade unions, or from within the hierarchy of party organizations. By the early 1990s, the prevalence and evils of seshu (hereditary suc- cession of electoral turf within bonds of kinship, typically from a father to his son) under the LDP regime became widely criticized (Inada 2009; Uesugi 2009). The media repeatedly questioned the quality of the MPs of the ruling parties, claiming that Japanese politics was hopelessly chaotic largely because it was filled with hereditary MPs, and the path to the Diet was closed to talented people who would have been able to offer solutions. With the electoral reform from SNTV to MA/FM in 1994, the parties faced new challenges in their nomination strategies. Under the old system where factions took responsibility for candidate recruitment (chapter 3, this volume), conservative contestants who failed to win an LDP nomina- tion often ran in the general election as independents. If they were able to win, the LDP usually allowed them to join the party. In effect, the LDP deferred difficult coordination decisions to the general election, allow- ing the electorate to sort things out. In the SMDs used in MMM, this strategy was no longer an option, since having two conservative candidates in the same district was a recipe for disaster. Thus, nominations became more critical, and party leaders were legitimized to take a more active role and exert more power in determining who would be nominated. The new opposition parties faced a different problem, that of finding enough quali- fied candidates to run in every district. The electoral reform increased the number of nominal districts from 129 to 300. Moreover, under the old system, even a locally weak party could hope to win one seat in most dis- tricts, and the possibility of winning attracted quality candidates. Under the new system, many of the districts were hopeless and thus unappealing to ambitious politicians. The difficulty in finding enough qualified candi- dates was exacerbated by the fact that opposition parties had never done well in prefectural assembly elections and did not have large numbers of established local politicians trying to move up to the national legislature 152 Mixed-Member Electoral Systems in Constitutional Context (Scheiner 2006). Electoral reform thus created a need for opposition par- ties to explore new strategies to find new pools of talent. With many vacant seats to fill, the opposition parties naturally were the pioneers to try out kobo.'7 In 1990, the JSP, then suffering from a shortage of candidates, was the first of the major national parties to recruit candi- dates for elected positions through kobo, although the practice was limited to the city/ward levels. It is widely recognized that the first political party to adopt kobo for national elections was the Nihon Shinto (People's New Party, 1992-94). It fielded three candidates recruited through kobo in the 1993 HR election. One of them was Yukio Edano, who later became a leading figure in the DPJ. The Shinshinto (New Frontier Party, 1994-97) used kobo extensively for candidate recruitment and managed to field can- didates in 235 SMDs in the first election under MA/MM in 1996. It was the first time the largest opposition party fielded candidates for over half of the district seats in an HR election since the JSP did so in the 1958 general election. The DPJ was the first major party to conduct kobo repeatedly over an extended period of time. The party conducted large-scale national kobo five times from 1999 to 2009. The LDP also tried out kobo on a much smaller scale as early as in 1994, but it did not fully embrace kobo until after the party's defeat in 2009.18 The precise institutional design of kobo differs from party to party, but they can be broadly grouped into two categories: centralized and decen- tralized. The national headquarters of the party manages the centralized type, and the prefectural branches and the district chapters play dominant roles under the decentralized type. The former became well known for its use by the DPJ, mainly for the elections for the HR from 2000 to 2009. Other relatively new third parties, such as the Minna no To (Your Party, 2009-), the Genzei Nippon (Tax Cut Japan, 2010-), and the Nohon Ishin no Kai (Japan Restoration Party, 2012-) also adopted similar forms of cen- tralized kobo. The latter was implemented widely by the LDP for the 2012 HR election. While the LDP adopted the decentralized version because it was more conducive to its entrenched local organization, parties with relatively short histories and less developed organization typically had no choice but to employ the centralized form. The initial goal of adopting kobo was to recruit potential candidates who the parties could not reach to through their traditional internal net- works and to find candidates to run in weak districts. This was especially the case with the DPJ. Gradually, however, insider aspirants started to use the DPJ kobo path to take advantage of its legitimacy. The LDP also Innovations in Candidate Selection Methods 15 3 used kobo to fill in weak districts, but its decentralized version of kobo, its entrenched local organization, and its more uniform application of kobo to open seats led the system to develop into a mediating device for intraparty competition. The DPJ Kobo System The current DPJ was formed in 1998 by absorbing three smaller parties19 into the old DPJ, which had been established in 1996. The members of the new DPJ came from different backgrounds: some of its leading figures originally belonged to the LDP, while others came from the Democratic Socialist Party, which had been established by the conservative wing of the old JSP. Right before the 2003 HR election, another major conservative force, the Liberal Party, dissolved, and joined the DPJ. As a relatively new and small party with underdeveloped local organizations, the DPJ suffered from an inadequacy of candidates, and this was the background for the adoption of kobo. Within the DPJ, incumbents were given priority over others and were renominated automatically in most cases. Candidates who lost in the previ- ous election were renominated once, but those who lost two elections in a row were not nominated for a third time. Only when the district had no incumbent or held-over candidate from the previous election did the party look for a new candidate. In such a case, the regular candidate selection process was initiated at the district level. The HR district chapter worked with the kenren (prefecture branch) in finding a candidate via the tradi- tional internal network. The kenren then asked the national headquarters for the endorsement of their choice. The national headquarters reserved the final say. When the district chapter and the kenren failed to field or agree on a candidate, the headquarters sent in a kobo-selected candidate. The SMDs where this happened were typically very weak. However, the party also attempted to use kobo as a public relations tool for enhancing its image by putting flashy ads in major newspapers with national circulation. The DPJ kobo was not conducted independently for each vacant dis- trict. Instead, the party held nationwide general kobo five times between 1999 and 2009.20 All of these were conducted in a similar manner, and each kobo supplied candidates to districts all over the country for the five HR elections for the Diet from 2000 to 2012.21 There were 74 kobo-selected candidates who debuted between 2000 and 2012 under the DPJ banner (table 5.5). The DPJ kobo for the Diet elections were directly managed by 154 Mixed-Member Electoral Systems in Constitutional Context the national headquarters. The headquarters had to take charge because the party used kobo to find candidates to run in vacant districts when the kenrens could not find anyone on their own. A kobo applicant had to pass three hurdles in order to become an offi- cial DPJ candidate: passing the kobo screening at the headquarters (Stage 1), securing an informal nomination for a specific district from the district chapter and the kenren (Stage 2), and getting this district-designated nomi- nation officially approved by the national headquarters (Stage 3). Tech- nically, the kobo system covered only the first stage, and both kobo and nonkobo candidates had to go through the latter two stages. The Stage 1 screening was handled by the national headquarters office. The applicants submitted curricula vitae and essays, and those who passed the initial document screening advanced to the in-person interview ses- sions. A few MPs joined the party staff in interviewing each applicant for about 20 minutes on what he or she wanted to do as an MP and in which district he or she wished to run. Interviewers focused on personality rather than policy orientation because the DPJ, being a hodgepodge of groups with different backgrounds, had no official policy positions it could enforce on its candidates. The total number of applicants for the DPJ national kobo increased from 564 in 1999 to nearly 2,000 in 2009. However, the odds of passing this first stage of screening remained constantly about one to nine. Those kobo-screenees who passed Stage 1 were eligible to advance to Stage 2-seeking district nomination. The headquarters served as an inter- mediary by matching these screenees and the kenrens with vacant districts. During the matching process, the applicants' personal roots in the districts or prefectures counted heavily. Finding districts to run in was often the toughest hurdle for the kobo screenees. Most of the better seats had been already taken, and the remaining seats often looked hopeless. Less than TABLE 5.5. Number of Kobo Candidates Who Ran and Won in HR Elections DPJ LDP Year Ran Elected Ran Elected 2000 17 3 0 0 2003 9 3 0 0 2005 18 1 26 22 2009 29 28 6 0 2012 1 0 83 76 Total 74 35 115 98 Source: Compiled by authors from party records. Innovations in Candidate Selection Methods 15 5 20% of these district-seekers were eventually embraced by district chapters and the kenrens and ran in elections as official party candidates. Stage 3, approval by the national headquarters, was not an automatic rubber stamp process. The person was not officially regarded as a DPJ candidate until after obtaining final endorsement from the headquarters. Although the national headquarters normally tried to respect local selec- tions and most of the local choices were endorsed, the headquarters would suspend approval or even replace candidates if kenren-chosen candidates looked too weak in the polls or did not seem to be campaigning hard enough. Overall, only 2% of the initial kobo applicants ended up running in the HR elections. We should note, however, that it is difficult to determine the quality of the applicants because the party never released any detailed information about the kobo applicants except for those who were officially nominated. The names of competitors had to be kept secret because run- ning for a public office in Japan almost always meant that one had to quit his or her current job before the election. No one wanted to quit before knowing whether he or she would get the party's nomination and could actually run in an election. Few people would apply for kobo if their names were not kept confidential. The use of kobo helped the DPJ prepare to take over power in 2009 by finding nontraditional outsider types of candidates to run in the weak districts. However, the party gradually started to ask those who used to be recruited through traditional paths to apply for kobo, as well, in order to add legitimacy to their nominations. For example, those members of pre- fecture assemblies who used to join the selection process at Stage 2 were often considered to have "stolen" the nomination by disgruntled kobo- qualified applicants. By the 2009 election, in order to avoid the impression of manipulation, the DPJ encouraged all candidates, including those from traditional career paths, to apply for kobo. Even though it was increasingly used to confer legitimacy on nominees, the kobo system was still fundamentally a nontransparent process. Deci- sions at each stage were made behind closed doors. The general public did not know exactly how the kobo selection process worked, who the kobo applicants were, which of them passed the screening stage or the match- ing stage, or what the bases were for the decisions at each stage. No offi- cial records were kept for the decision making at the matching stage. The only way the voters knew that the candidates were selected through kobo was through the label of kobo attached to them when they were presented as the nominated candidates. This opaque process allowed national and 156 Mixed-Member Electoral Systems in Constitutional Context local faction bosses to exert influence behind closed doors during the selec- tion processes. As a result, the DPJ kobo came to function more as a black box convenient for various actors within the party in contrast to its public image as an open and fair CSM (Shoji 2013). The LDP Kobo System The earliest kobo practice by the LDP was observed in 1994 by the Ishikawa kenren for the 1995 House of Councillors election. After the 1994 electoral reform, several kenrens tried kobo in some districts in preparation for the 1996 HR election. Yet the LDP did not widely conduct kobo for national offices until it suffered heavy losses in the 2003 House of Council- lors election. Encouraged by a few kobo successes in special elections, the LDP conducted kobo in about 20 districts before it conducted a special emergency national kobo right before the 2005 HR election. For the latter, 868 applications were submitted in four days. Twenty-four kobo-selected candidates ran in the 2005 HR election, and twelve won in the SMDs. A further nine lost their district race but were elected on the PR list (Asano 2006). For the 2009 HR election, the LDP nominated only six candidates through district-level kobo.22 After the defeat in the 2009 HR election, the LDP attracted fewer candidates, and the party started to use kobo exten- sively to fill in the vacant districts. Facing new challenges as an opposition party and continuing pub- lic complaints over seshu, the LDP made it mandatory to conduct kobo for any open HR district. Right after the 2009 HR election, the then- opposition LDP was no longer as attractive to ambitious new politicians and had some difficulties in recruiting satisfactory candidates through its traditional paths. In response, the LDP once again turned to kobo. As the popularity of the party recovered, the competition for LDP nominations intensified, and the LDP responded by further institutionalizing the use of kobo. In this section, we focus on the LDP's practices in preparation for the 2012 HR election. The party heavily protected its incumbents and renominated them automatically, as it had done throughout its history. After the 2009 HR election, the candidates who lost were automatically renominated if they were 65 years old or younger and had lost in the SMD but won at least 70% as many votes as the district winner. The party only began the can- didate selection process when there was neither an incumbent nor a held- over candidate in the district. Although the national headquarters had the final say, the candidate Innovations in Candidate Selection Methods 157 selection processes of the LDP were bottom-up, starting at the district level. A district chapter worked with its kenren in recruiting and selecting a candidate. Because the party made it a rule to hold kobo for any open district nomination after the 2009 HR election, kobo was conducted not just in weak districts but also in strong districts such as those where strong incumbents had just retired. During this period, the party attempted to bar the seshu practice, but it had to give up after facing strong resistance. The party justified this by insisting that anyone, seshu or not, who was selected through kobo was fairly chosen and thus qualified. The LDP fielded 83 candidates through kobo for the 2012 HR election. The largest difference from the DPJ kobo was that the LDP kobo was always conducted at the prefecture or district level. In most cases, each kobo was held specifically for a certain district, and the kenren would determine the specific kobo procedures. Thus, there was a wide variety in the LDP kobo practices observed across the country. Typically, a selec- tion committee of 20 or fewer people was set up at the district or kenren level for each kobo. Outside experts or prominent local figures represent- ing the interests of local industry were often invited into the committee. The selection processes normally entailed paper-screening and interview sessions. Applicants were always asked to submit essays, and, in a majority of cases, the public announcement listed some connection to the district as a prerequisite or at least as highly desirable. Complete outsiders with no connection to the district or the prefecture were usually not welcome. In some cases, the selection committee required applicants to fill in survey sheets on policy preferences. The LDP witnessed the DPJ's tra- vails stemming from the lack of policy coherence within the party and attempted to avoid the same problem by screening the ideological lean- ings of kobo candidates. The party occasionally asked applicants to sign a pledge that they would not run against the LDP nominee if they were not nominated. While most of the kobo selection took place in closed-door meetings, there were some instances in which the applicants were asked to compete by giving speeches in public or in which the party used opinion polls to identify the most attractive candidate. There were even some kobo cases where party members voted to decide the outcome. These closed primaries did not necessarily lead to more open or fair selections, however, since the competitors with a longer history of local party activities had usu- ally cultivated more supportive members and the newcomers were put at a severe disadvantage. Broadly speaking, there were three observed patterns in the actual com- petition under the LDP kobo. The first was districts with a single likely 158 Mixed-Member Electoral Systems in Constitutional Context winner, such as those that would previously have been determined by seshu. Under this category, kobo functioned to justify the predetermined outcome. The second category included districts with two or three major contestants. Those contestants were often well-established local politi- cians in the district, and the party had a hard time mediating a compromise among them. Kobo helped to justify the final outcome and persuade the losers to accept the result. In this case, kobo worked as a vehicle of conflict resolution. The third category included districts with no strong competi- tors known prior to kobo. These cases typically involved districts where the party was extremely weak and potential quality candidates were reluc- tant to run. The party had to recruit complete amateurs from outside the internal network who would dare to run against the strong incumbents of another party. Discussion Although they were called with the same name, kobo, and they shared a similar function at the early stage of fielding candidates in the weak dis- tricts, the centralized and decentralized versions adopted by the DPJ and the LDP, respectively, evolved in very different ways. The contrast is sum- marized in table 5.6. The headquarters-led DPJ kobo process remained extremely nontransparent in its selection criteria and the decision-making protocol. At the district assignment stage, where the negotiation process remained completely informal and hidden, anything was possible. Not unaware of these drawbacks, many DPJ staff and MPs testified they never perceived kobo as an ideal CSM. The centralized form of kobo left parties vulnerable to collapse in more difficult times. New parties had to use it because they had underdeveloped party organization and suffered an inadequate supply of candidates. In turn, this dependence on kobo led to negligence in building internal recruitment and training systems. The TABLE 5.6. Characteristics of DPJ and LDP Kobo DPJ kobo LDP kobo Unit of kobo At-all (national) Mostly by district Implementation of kobo Headquarters Prefectures/districts Scope of participation Very limited Relatively broad Final say Headquarters Headquarters Period of intensive kobo practice For 2000-2009 HR elections For 2012 HR election State of party Opposition Opposition Primary initial motivation Filling vacancy Improving public image Use of closed primary None Several Innovations in Candidate Selection Methods 159 DPJ kobo was used by ambitious aspirants who wanted to become MPs while bypassing an unappealing apprenticeship in local politics. These can- didates had no loyalty to the party and were prone to leave the party when they did not like the policy decisions made by the leaders. The number of kobo applicants plummeted as the popularity of the party dropped, which revealed that how well kobo functioned depended heavily on whether the party's electoral prospects looked promising. The quick rise and fall of the Japan Restoration Party, which was completely dependent on the charisma of a single leading figure and an extensive use of centralized kobo, illus- trates both the advantages and dangers of this CSM. The structure of cen- tralized kobo left the parties with little space for improvement in terms of intraparty democracy. It merely opened the door to the smoke-filled room to outside aspirants, but the decision-making processes were left exclusive and nontransparent. The decentralized kobo that the LDP experimented with presented different possibilities. The kenrens tried different ideas, and good examples were copied by others. Some examples of policy debate in public, formal use of polling surveys, and even closed primaries were observed in the course. While many kenrens preferred to remain secretive and exclusive about nomination processes when possible, the LDP experience suggests that kobo might serve as a venue for CSM democratization if implemented by a party with strong local party organization. Although much of the decision-making processes of the LDP kobo remained nontransparent to most voters, the institutionalization of kobo that took place at the prefec- ture and district levels helped improve the fairness and legitimacy of the nominations. The LDP kobo reform is even showing some signs of being nonreversible. With an increase in the kobo practices, it has become harder for the party to reject holding kobo when demanded by competitors. While it is hard to measure the impact of kobo on factions within the DPJ, which started to use kobo right after its inception, it should be fair to say that kobo did not help factions within the LDP. If the introduction of SMD had already made things harder for the factions (chapter 3, this volume), opening competition to outsiders in each district made it almost impossible for the factions to coordinate across districts and prefectures. Some studies investigate the impact of kobo on the type of candidates who won nominations. Smith, Pekkanen, and Krauss (2013) examined DPJ nominations through the 2009 HR election and found that kobo nomi- nees tended to be lower quality on average than nonkobo nominees. DPJ kobo candidates tended to have less connection to the district, and aside from the female-only kobo in 1999, kobo was actually less likely to recruit 160 Mixed-Member Electoral Systems in Constitutional Context female candidates. Overall, the kobo candidates had a higher probability of winning, but the impact of kobo disappears after controlling for the party strength in the district and candidate quality. As for the LDP kobo, specifically in the run-up to the 2012 HR election when kobo became the default CSM for the party, Smith (2013) argues that it did not significantly change the backgrounds of candidates who were nominated. Kobo helped to push down the average age of first-time nomi- nees, but the backgrounds of those kobo candidates remained largely simi- lar to the traditional pools: local politicians, MP aides, and national-level bureaucrats. We should also note that, although decreasing in number over time, seshu was not wiped out by kobo. Conclusion This chapter has analyzed new innovations in CSMs for legislative elec- tions in Taiwan and Japan. While the origins of the polling primary in Taiwan and kobo in Japan both predate electoral reform, we argue that the shifts from SNTV to MMM created pressures for parties in both coun- tries to intensify the use of these two institutions. Further, the effects of the changing electoral rules on CSMs need to be addressed in the con- text of the different constitutional systems of the two countries. Specifi- cally, Taiwan's semipresidential system deemphasizes legislative seniority while Japan's parliamentary system always relies on senior MPs to form the government. Such contextual variance leads to the different purposes and scopes of the CSM reforms adopted by major parties in the two countries. In Taiwan, where seniority is less valuable and incumbents are not pro- tected, the parties converged on the use of the polling primary in strong districts. The polling primary has proven useful to the parties by success- fully mediating intraparty conflict, and it is also beneficial to the general public to the extent that it is a disincentive to corruption. In Japan, where seniority is more valuable and incumbents are heavily protected, both major parties adopted kobo in weak districts. This allowed them access to a wider pool of potential candidates and helped to defuse charges that insid- ers were illegitimately monopolizing nominations. Of course, the CSMs continue to evolve. In Taiwan, many worry that the polling primary harms intraparty democracy and party-building by leaving no role for party members to influence nominations. After the 2012 elections, there was discussion in both parties about reinstating some mix of closed primaries and polling primaries. Though neither party has opted Innovations in Candidate Selection Methods 161 for a change from the 2012 rules as of this writing, the parties continue to search for a CSM that will give party members a bigger role while still resolving conflict and discouraging corruption. In Japan, there are serious concerns about the effects of centralized kobo, which has proven useful for young and popular parties but has also hindered efforts to develop party institutions that can survive when a party's fortunes decline. The decentral- ized version may prove more sustainable. It shows signs of taking root and has demonstrated some potential for intraparty democratization. NOTES 1. In the new electoral regimes in both Taiwan and Japan, the nominal tier plays the dominant role in shaping incentives. In both countries, a relatively large pro- portion of seats are determined in single-member districts. Further, in Japan the practice of dual candidacy results in SMD candidates taking up most of the space on PR lists, thus leaving relatively few spots for PR-only candidates. This chapter thus focuses on the linkage between electoral rules for the nominal tier and CSMs. Nemoto and Tsai take up the topic of list tier nominations in chapter 6, this volume. 2. This has been less of a problem in Taiwan than in Japan for three main rea- sons. First, Taiwan has only 73 SMDs compared to Japan's 300. Parties simply have fewer districts to worry about. Second, as part of Taiwan's electoral reform, the total number of seats was cut in half. That meant that Taiwan had large numbers of incumbents chasing a relatively small number of seats. Third and most important, like the LDP and unlike the DPJ, both of Taiwan's major parties were well estab- lished and had already amassed reservoirs of ambitious potential candidates within their ranks. 3. DPP cadres include party members of the Legislative Yuan, National Assem- bly, Taiwan Provincial Assembly, as well as Taipei and Kaohsiung city councils. 4. Both parties utilize more centralized methods to determine their candidates for party list vote. For more on party list nominations, see chapter 6 by Nemoto and Tsai in this volume. 5. Officially, the KMT nomination rules still include the clause allowing party members' votes to account for 30% of the decision. However, aspirants may agree to skip this stage, and in 2012 all KMT aspirants chose to use only polling results to determine nominations. 6. Interview with Lee Hung-chun, November 16, 2011. 7. Interview with Lin Shu-fen, November 16, 2011. 8. In-house sampling refers to a protocol used to determine which member of the household should respond to the survey. Typically, the interviewer determines how many adult males and females live in the residence. Based on these numbers and a predetermined schedule, the interviewer will ask to speak to, for example, the second-oldest female. On in-house sampling, see Hung 1996. 9. Interview with deputy secretary general Hung Yao-fu, November 14, 2011. 10. The interparty comparison was used when the DPP candidate's support was higher than the KMT candidate's. Where the KMT candidate was more popular 162 Mixed-Member Electoral Systems in Constitutional Context than any of the DPP candidates, an intraparty head-to-head question was used. However, most of the districts that required a polling primary were those in which the DPP was more popular, so we consider the interparty comparison to be the default. 11. The DPP finished its nomination process before the KMT, so the KMT can- didate was not always known. However, because of the KMT's landslide victory in 2008, most districts, including many of the DPP's stronger districts, had a KMT incumbent that the DPP assumed would be renominated in 2012. 12. The KMT intentionally waited until the DPP had completed its nomination process so that it could choose the best candidate to match up with that specific DPP opponent. (Interview with a member of the KMT's think tank, Dr. Hsieh Hsian-chin, November 16, 2011.) 13. Interview with deputy secretary general Hung Yao-fu, November 14, 2011. 14. The fear over vote-buying was an important factor in the DPP's decision to move to a 100% polling primary in 2012. In November 2007, the KMT-dominated legislature revised the election law and made it illegal to buy votes in party prima- ries, not just in the general election. The DPP worried that the KMT would use its control over the bureaucracy to subject its nominees to intense scrutiny and might be able to disqualify some of them. Rather than fight the election with a roster of tainted candidates, the DPP simply eliminated the party-member voting section of its CSM. 15. Many people also worry that their party's CSM is vulnerable to manipula- tion by the other party. For example, KMT supporters might participate in the DPP's polling primary and express support for the weakest DPP candidate. These fears are probably overblown, and DPP deputy secretary general Hung Yao-fu even identified the low vulnerability of the polling primary to manipulation from the other party as one of its major advantages (interview, November 14, 2011). On the one hand, the closed party primaries were probably more vulnerable since the other party could strategically launch a registration drive and send large number of its supporters to vote in the other party's primary. In contrast, since polls are random, this action would need to be individually undertaken by large numbers of other- party supporters who would further have to agree on who the weakest contestant was. The polling primary simply makes the coordination costs very high. On the other hand, anecdotal evidence suggests that the same people who are worried about the other party manipulating their nominations tend to be disinterested in manipulating the other party's nominations. 16. A common sarcastic jab is that DPP no longer stands for Democratic Pro- gressive Party (minjindang); now it stands for Democratic Polling Party (mindi- aodang). 17. A nonpartisan civic movement emerged in 1990 to use kobo to cultivate and promote fresh candidates without party stigma for local executive offices. The movement originated in the mayoral election in Asahikawa City, Hokkaido, and it gradually spread to other prefectures around the country over the next few years. The popularity of this movement and the media attention it drew might have had some impact on the decision of parties to try kobo. 18. No party has used kobo for the selection of PR-only candidates. 19. The three smaller parties were Minsei To (Good Governance Party, 1998), Innovations in Candidate Selection Methods 163 Shinto Yuai (New Fraternity Party, 1998), and Minshu Kaikaku Rengo (Demo- cratic Reform Party, 1989-98). 20. In addition to these general ones, the DPJ headquarters conducted a female- focused kobo in 1999 and a district-specified kobo in 2005. The latter was special in the sense that vacant districts were listed in advance; they had remained available because they were all extremely difficult for the DPJ. 21. We located in newspapers nine cases of kobo conducted by the DPJ kenrens at the prefecture level for the HR elections between 2000 and 2009, and there might have been more. However, these cases were not recognized by the national headquarters, and no systematic records have been kept for them. For this paper, we focus on the national kobo conducted by the headquarters. 22. The number of kobo candidates differs across studies due to the way of count- ing. More than a few cases of kobo were observed that were called kobo by the ken- rens but were not recognized by the national headquarters. Here in this chapter, we focus on those that were officially recognized by the national headquarters. REFERENCES Asano, Masahiko. 2006. Shimin Shakai ni Okeru Seido Kaikaku: Senkyo Seido to Kouhosha Rikuruto [Institutional reform in civil society: Electoral system and candidate recruitment]. Tokyo: Keio University Press. Browne, Eric, and Dennis Patterson. 1999. "An Empirical Theory of Rational Nominating Behavior Applied to Japanese District Elections." British journal of Political Science 29:259-89. Cheng, Tun-jen. 1989. "Democratizing the Quasi-Leninist Regime in Taiwan." World Politics 41 (4): 471-99. Cox, Gary W., and Emerson Niou. 1994. "Seat Bonuses under the Single Nontrans- ferable Vote: Evidence from Japan and Taiwan." Comparative Politics 26:221-36. Cox, Gary W., and Frances Rosenbluth. 1993. "The Electoral Fortunes of Legisla- tive Factions in Japan." American Political Science Review 87 (3): 577-89. Cox, Gary W., and Michael E Thies. 1998. "The Cost of Intraparty Competition: The Single Non-transferable Vote and Money Politics in Japan." Comparative Political Studies 31 (3): 267-91. Gallagher, Michael, and Michael Marsh, eds. 1988. Candidate Selection in Compara- tive Perspective: The Secret Garden of Politics. London: Sage. Hazan, Reuven Y., and Gerrit Voerman. 2006. "Electoral System and Candidate Selection." Acta Politica 41:146-62. Huang, Teh-fu. 1996. "Elections and the Evolution of the Kuomintang." In Tai- wan's Electoral Politics and Democratic Transition: Riding the Third Wave, ed. Tien Hung-mao. Armonk, NY: M. E. Sharpe. Hung, Yung-tai. 1996. Huzhong Xuanyang zhi Yanjiu [In-house sampling]. Taipei: Shihying. Inada, Shigeru. 2009. Seshu Giin: Kozo to Mondaiten [Hereditary MP: Background and problems]. Tokyo: Kodansha. Kasapovic, Mirjana. 2001. "Nominating Procedures in Democratic Polities." En: PolitickalMisao 38 (5): 3-17. Krauss, Ellis S., and Robert Pekkanen. 2004. "Explaining Party Adaptation to Elec- 164 Mixed-Member Electoral Systems in Constitutional Context toral Reform: The Discreet Charm of the LDP?"Journal ofapanese Studies 30 (1): 1-34. Krauss, Ellis S., and Robert Pekkanen. 2011. The Rise and Fall ofapan's LDP: Politi- cal Party Organizations as Historical Institutions. Ithaca: Cornell University Press. Lundell, Krister. 2004. "Determinants of Candidate Selection: The Degree of Cen- tralization in Comparative Perspective." Party Politics 10 (1): 25-47. Norris, Pippa. 1997. Passages to Power: Legislative Recruitment in Advanced Democra- cies. Cambridge: Cambridge University Press. Ranney, Austin. 1981. "Candidate Selection." In Democracy at the Polls: A Compara- tive Study of National Elections, ed. David Butler, Howard R. Penniman, and Aus- tin Ranney, 75-106. Washington, DC: American Enterprise Institute. Scheiner, Ethan. 2006. Democracy without Competition in Japan: Opposition Failure in a One-Party Dominant State. Cambridge: Cambridge University Press. Shoji, Kaori. 2013. "DPJ Kobo on the Ground: Opening Candidate Nominations without Democratization" [in Japanese]. Gakushuin Diagaku Hougakkai Zasshi [Gakushuin Review of Law and Politics] 49 (1): 181-210. Shomer, Yael. 2012. "What Affects the Candidate Selection Process? A Cross- National Examination." Party Politics, published online, April 11. http://dx.doi. org/10.1177/1354068811436060. Smith, Daniel M. 2013. "Candidate Recruitment for the 2012 Election: New Par- ties, New Methods. . . Same Old Pool of Candidates?" In Japan Decides 2012: The Japanese General Election, ed. Robert Pekkanen, Steven R. Reed, and Ethan Scheiner. Hampshire, NY: Palgrave. Smith, Daniel M., Robert J. Pekkanen, and Ellis S Krauss. 2013. "Building a Party: Candidate Recruitment in the Democratic Party of Japan, 1996-2012." In'apan under the DPJ: The Politics of Transition and Governance, ed. Kenji E. Kushida and Phillip Y. Lipscy. Stanford: Walter H. Shorenstein Asia-Pacific Research Cen- ter. Uesugi, Takashi. 2009. Seshu Giin no Karakuri [Tricks of hereditary MPs]. Tokyo: Bungei-Shunju. Wang, Yeh-Lih. 2006. Bi/iao Xuanju Zhidu [Comparative electoral systems]. Taipei: Wunan Book. Wu, Chung-li. 2001. "The Transformation of the Kuomintang's Candidate Selec- tion System." Party Politics 7 (1): 103-18. Wu, Chung-li. 2008. Zhengdang yu Xuanju: Lilun yu Shijian [Political parties and elections: Theory and practice]. Taipei: Sanmin. Wu, Chung-li, and Dafydd Fell. 2003. "Taiwan's Primaries in Comparative Per- spective." Japanese Journal of Political Science 2 (1): 23-45. SIX Post Allocation, List Nominations, and Preelectoral Coalitions under MMM Kuniaki Nemoto and Chia-hung Tsai A theory of electoral systems (Shugart 2001) predicts that the move from SNTV to MMM should produce a trend away from personal politics and toward party-centered politics. The literature on the recent elections in Japan suggests that party labels and the leadership's popularity now signifi- cantly influence candidates' electoral fortunes (McElwain 2012; Patterson and Maeda 2007) and clientelism is now being replaced by programmatism (Rosenbluth and Thies 2010). Furthermore, in 2009 Japan had a govern- ment turnover through elections for the first time in more than 60 years, because the combination of the DPJ's potential and disillusionment with the LDP government's policy performance convinced most swing voters: a poll on the 2009 turnover, for example, showed that 61.2% of voters thought party reputations were important in making voting decisions, compared to 29.0% who considered individual candidates' personal traits as important (Akarui Senkyo Suishin Kyokai 2010). Reed, Scheiner, and Thies (2012) argue that the swings in Japan's recent elections are built on coherent party images and in particular parties' programmatic policy manifestoes, showing that "candidates' party affiliation mattered more for their chances of victory than their individual resumes" (373). This record high importance attached to party label is a flip side of the fact that the importance of individual candidates has been declining over time. Surveys conducted in the mid-1960s found 60.4% of the vot- 165 166 Mixed-Member Electoral Systems in Constitutional Context ers in Japan voted on the basis of the candidate, compared to 43.3 % in the United States, 32.5% in Canada, 23.8% in Germany, and 5.4% in the United Kingdom (Rochon 1981). This candidate-based voting was domi- nant until the 1993 elections, the last elections before the electoral reform, with nearly a half (49.1%) of the voters still voting based on the candidate (Akarui Senkyo Suishin Kyokai 2010). Against this background, this chapter investigates how parties strategi- cally adapted themselves to the new electoral environment. Specifically, it compares what parties do between and during election periods in order to maximize votes and seats, how this has changed compared to the prereform SNTV period, and how the two countries differ. A simple prediction is that parties should concentrate their efforts on appealing to voters with their policymaking abilities and attractive public policy pledges. Beyond this simple prediction, however, there should be many more vote mobilization strategies that parties can adopt. Their choice from among these strategies should depend on the institutional contexts in which parties compete and the institutional resources available for the party leadership to utilize. The basic argument throughout this chapter is that, as articulated by Batto and Cox in their introductory chapter, the transition from SNTV to MMM should simplify the intraparty bargaining process for leaders to utilize such resources. This is because, compared to SNTV, which systematically generates intraparty fragmentation and fac- tionalism (Grofman 1999), MMM increases the importance of party repu- tations, policy platforms, and leaders' popularity in electoral competition (Shugart 2001). Consequently, MMM centralizes the power of the party leadership against intraparty factional bosses, to the extent that parties should use available tools and resources more strategically to maximize votes and seats. Among the many strategies, we focus on three areas: post allocation, list nomination, and candidate coordination. Along with the main argument briefly articulated above, party leaders should be now expected to (1) allo- cate cabinet positions and other privileges more autonomously from intra- party factions to members; (2) nominate candidates with clean and fresh images at the top of party lists, such as women and professionals who are widely known across the country; and (3) form preelectoral coalitions in the nominal SMD tier to solicit small party supporters. We generally find that parties in both Japan and Taiwan are moving toward these expected directions, suggesting that they are indeed strategically adapting them- selves to the new electoral environment. However, we also argue that nuanced but very important rules Post Allocation, List Nominations, and Preelectoral Coalitions 167 the constitution and the dual candidacy with the best loser provision institutionally constrain parties' strategies to the extent that these naive expectations do not hold under certain contexts. First, compared to a par- liamentary system, where appointing legislators to cabinet positions serves as institutional bonding for the cabinet to maintain the confidence of the legislature, semipresidentialism implies much more independence between the president and ruling party members. Therefore in Taiwan, promoting and demoting legislators cannot be as effective as in parliamentary Japan. Second,Japan's somewhat unique "best loser provision"-dually nominated candidates can have the same ranks and their actual ranks are determined by their performance in their SMDs-makes it difficult for leaders to give favorable positions to quality candidates. Rather, leaders now simply put almost all the SMD candidates on a level playing field. Meanwhile, the absence of such a provision enables Taiwanese leaders to be more strategic in selecting and rank-ordering list candidates. Third, the presence of the list tier, on which small parties rely for their electoral survival, makes can- didate coordination rather complicated in mixed-member systems. Large parties in parliamentary Japan can credibly commit not to exploiting small parties by appointing ministers and forging coalition agreements, but the presence of presidential coattails and interbranch independence in Taiwan can make coordination between parties for candidate entry rather difficult. This chapter is organized in three sections. First, we will discuss how parties allocate posts for electoral purposes in Japan and Taiwan. Second, we will discuss how they nominate candidates on the list tier, where par- ties' images and reputations matter greatly. Third, we will investigate how they coordinate candidates and form preelectoral coalitions to maximize the chance of winning SMD seats. Post Allocation Strategy This section focuses on how parties strategically allocate posts in the gov- ernment to incumbents for electoral purposes. As Kobayashi and Tsuki- yama show in chapter 3, prior to reform SNTV's intraparty fragmentation effects made the bargaining process for post allocation rather complicated. Since factional heads vied for the limited number of lucrative positions in the government, the prime minister needed to allocate cabinet positions proportionally to each of the factions according to their sizes (Ramseyer and Rosenbluth 1993). In addition to this norm of the factional balance (habatsu kinko), promotion of backbenchers to legislative committee chairs, 168 Mixed-Member Electoral Systems in Constitutional Context vice ministers, and ministers was strictly determined based on seniority (nenko joretsu). These party rules were effective methods to contain ambi- tious factions and keep the party united (Nemoto, Krauss, and Pekkanen 2008), but significantly constrained the leadership. The transition to MMM should simplify this intraparty game for two reasons: first, factional heads' influence should become weaker; and second, the leadership can exercise greater influence based on the ever- increasing importance of party labels and leaders' image. Thus, the lead- ership should more freely and strategically target certain members when giving out the government's privileges. Thus Kobayashi and Tsukiyama show that the prime minister's faction now receives disproportionately more posts in the cabinet. This section further explores how this centralized control over posts now enables the leadership to utilize post allocation for electoral pur- poses, in line with Pekkanen, Nyblade, and Krauss (2006). They show that marginal members, such as zombies who were defeated in SMDs but still elected by PR, were more likely to assume posts connected to private and particularistic goods, so that they could improve name recognition in their districts and return to the Diet in the next election. They also show that senior and famous figures were assigned to high-policy posts, such as min- isters for security and treasury, so that the party could improve its policy- making reputations. However, the temporal trend shown in figure 6.1 suggests that parties give fewer positions to electorally marginal members (i.e., zombies) over time. Rather, being elected on the SMD tier has become nearly a necessary condition for a member to become a full minister since 2005. Zombies and PR members are now less likely to assume lower-level positions (vice min- isters and legislative committee chairs) than in 1996, 2000, and 2003. The DPJ governments avoided electorally marginal members in assigning new Diet committee chairs and party posts after the 2009 election: the then secretary-general of the DPJ, Ichiro Ozawa, prioritized SMD winners and made it clear that the party wouldn't give important posts to those elected on PR (Asahi Shimbun, October 15, 2009, 4). Rather, he ordered first-term members and zombies to go back to their local districts often and expand their supporter bases (Asahi Shimbun, October 30, 2009, 4). This appears to be somewhat incompatible with what Pekkanen, Nyblade, and Krauss (2006) find, especially given that now Japanese parties make the most use of dual candidacy and there are more and more zombies that the leadership needs to take care of. We suggest that Pekkanen, Nyblade, and Krauss (2006) might have Post Allocation, List Nominations, and Preelectoral Coalitions 169 Ministers 1996 Vice Ministers Diet Committee Chairs Ministers 2000 Vice Ministers Diet Committee Chairs Ministers 2003 Vice Ministers Diet Committee Chairs Ministers 2005 Vice Ministers Diet Committee Chairs Ministers 2009 Vice Ministers Diet Committee Chairs Ministers 2012 Vice Ministers Diet Committee Chairs I- I -a 0 20 40 60 80 100 percent SMD Zombie PR Fig. 6.1. Lower-house members in government and Diet posts by different types ignored two things. One is a counterintuitive consequence from their model: if you want to increase the chance of getting good positions, you need to lose in the nominal tier. Thus, appointing marginal members to good positions should result in the problem of moral hazard, although it is somewhat unrealistic to assume that some SMD candidates might want to lose. The second thing is the change that the electoral reform could bring about: the decline of particularism and clientelism (Noble 2010). Thus, over time, the idea of having backbenchers use posts to allocate subsidies and other private goods to their districts should have become untenable, especially given the shrinking budgets of the Japanese government. These factors combined lead us to suspect the leadership gradually changed its appointment patterns from posts as "carrots to marginal mem- bers" to posts as "sticks to tame backbenchers." We argue that, as party reputations become more important than ever, the leadership now needs able and efficient ministers to lead bureaucrats, respond to tough questions from the opposition in the Diet, and implement the party's policy plat- forms. The LDP leadership gradually appointed loyal and able lawmak- ers, while intentionally ignoring the factional balance and seniority norms (Nihon Keizai Shimbun, September 19, 2006, 14; also see Nemoto, Krauss, 170 Mixed-Member Electoral Systems in Constitutional Context and Pekkanen 2008). Therefore, it could be hypothesized that, as long as posts in the government and the Diet are valued by members in getting attention from the media and exerting influence over the policy-making process, the leadership can encourage backbenchers to work harder in their local districts by appointing more SMD winners to these posts. This whipping strategy should look more appropriate under MMM than the compensation strategy. As one rough test for this hypothesis, we divided LDP members in 2000-2009 into four groups according to their SMD performance in two consecutive elections (1996-2000, 2000-2003, and 2003-5): (1) those who won SMDs in the first elections but won in the second as zombies (SMD- Zombie); (2) those who won as zombies for the two consecutive elections (Zombie-Zombie); (3) those who won as zombies in the first but won SMDs in the second (Zombie-SMD); and (4) those who won SMDs continuously (SMD-SMD). If Pekkanen, Nyblade, and Krauss's (2006) compensation story is correct, we should find more members promoted in the SMD- Zombie category. Meanwhile, if the leadership instead uses promotion to whip lazy members and reward hard workers, then those who lost their seats (SMD-Zombie) should be punished and those who regained their seats (Zombie-SMD) should be promoted. Here, in line with the literature (Ramseyer and Rosenbluth 1993, 86), we coded a member's promotion as his/her moving up the career ladder of (I) no post; (II) vice minister; (III) Diet committee chair; (IV) minister; and (V) second-time (or higher) min- ister.1 A note of caution is required in interpreting the results in figure 6.2, as by its nature the sample here excludes those who didn't run in the nominal tier for the two consecutive elections, and most (87.0%) of the 508 observations are concentrated in the upper-left "SMD-SMD" cell. Interestingly, the figure suggests that, contrary to the compensation story, the "SMD-Zombie" category in the upper-right cell is the least likely to get promoted. This implies that they are punished for losing their seats. On the other hand, consistent with the whipping story, the "Zombie-SMD" category in the lower-left cell is the most likely to get rewarded, suggesting that those who ousted opposition district incumbents were seen as making the biggest contributions to the party. The lower-right "Zombie-Zombie" cell looks slightly puzzling as nearly half of them were promoted, but six out of the 11 in this entry were in fact "favorably ranked" (discussed in the next section) hopeless candidates who continuously challenged invincible opposition party leaders. Excluding these loyal hard workers, only one (20%) of the five in the "Zombie-Zombie" category was promoted. Moving on to the case of Taiwan, we expect that in contrast to parlia- Post Allocation, List Nominations, and Preelectoral Coalitions 17 1 Second Election Won SMD Zombie "SMD-SMD" "SMD-Zombie" Won SMD 34.3% (152/443) 14.3% (2/14) First Election "Zombie-SMD" "Zombie-Zombie" Zombie 55.0% (22/40) 45.5% (5/11) Fig. 6.2. LDP members' promotion, 2000-2009 mentary Japan, the same logic should not hold under the semipresiden- tial regime, as Lin shows in chapter 2. In parliamentary Japan, the stakes are high for backbenchers to win their seats and expand the party's size in the legislature, as it will directly increase their chance of becoming ruling party members and getting promoted in the government. In other words, legislative confidence results in a much closer alignment between appoint- ments and vote-seeking incentives. Meanwhile, even though the Taiwanese president may be able to appoint legislators as ministers, he or she does not really have to do so, as he or she is independently elected by the people and his or her survival does not rely on legislative confidence. And legislators' reelection incentives should only be weakly tied to their office-seeking incentives-when they are appointed as ministers, they have to resign their legislative seats, which are won independently from the president. However, some recent anecdotes suggest that a somewhat similar story applies to the Taiwanese context. For example, a former cable TV news anchor, Chen Yi-chen, was picked directly by the KMT's central headquar- ters to run in Chiayi, one of the DPP's strongholds in southern Taiwan 172 Mixed-Member Electoral Systems in Constitutional Context (Taipei Times, August 8, 2011, 3). She also helped President Ma Ying-jeou organize his campaigning office in the south (Taipei Times, August 21, 2011, 3) and reportedly, rather than winning seats in the Legislative Yuan, Chen and the other KMT candidates running in the opposition's strongholds were expected to shore up support for Ma in the presidential election to be held on the same day (Taipei Times, September 20, 2011, 3). Although Chen lost, Ma still valued her efforts highly, as they helped the party deepen its ties with supporters (China Post, January 20, 2012), and he appointed her to be National Youth Commission minister after the election (China Post, February 1, 2012). In total, after the 2012 legislative elections, the KMT used cabinet positions as carrots when it rewarded five losers who had run against very strong opponents by appointing two as ministers, one as a deputy minister, one as head of a party branch, and one as spokesperson for the Executive Yuan. In addition to compensating losers with positions in the cabinet, the KMT is also using these positions to develop winnable candidates to com- pete with the DPP, especially in southern Taiwan. Reportedly the Executive Yuan was to restructure the administration by the end of 2012 and planned to have Chen Yi-chen, the aforementioned minister appointed right after the 2012 elections, lead a regional branch office in Chiayi City (Taipei Times, June 26, 2012, 1). The DPP criticized the KMT government over the move, as she was rumored to be preparing to run for the 2014 mayoral election in the city, a Pan-Green stronghold. The DPP even speculated that she might leverage the regional administration position to improve her local name recognition for the upcoming election (Taipei Times, July 14, 2012, 3). Similarly, when Lin Yi-shih had to step down as minister because of corruption charges, the media reported that it was a major loss for President Ma, who had been grooming Lin, a southern Taiwan native, as the KMT's candidate in the next Kaohsiung mayoral election. Lin had been given him the positions of head of the party's youth corps, vice chair of the party, chief executive of the party's Policy Coordination Committee, and cabinet secretary-general (China Times, June 30, 2012, 4). The flip side of all this is that the leadership can use cabinet appoint- ments as sticks: it can demote members when they do not conform to its expectations. For example, in the 2008 elections, Wang Yu-ting repre- sented the KMT in Tainan's 1st District, one of the DPP's strongholds, and lost. In exchange for her campaigning efforts, the president gave her the cabinet-level position of chairperson for the National Youth Commis- sion, with an eye toward using the Tainan native again in future elections to expand the party's support bases in the south. The chance came in 2011, Post Allocation, List Nominations, and Preelectoral Coalitions 17 3 when a seat in Tainan was made vacant and a by-election was scheduled, but she chose not to accept the KMT's request. The KMT's secretary- general, King Pu-tsung, said, "[She was] nurtured by the party, but refused to step up when needed.. . . A warrior does not have the right to choose the battles" (China Post, January 18, 2011), implying the party would punish Wang for her insubordination. Together with another KMT cabinet mem- ber, Kao Su-po, who also declined to go back to his district and run in the by-elections, Wang was summarily fired from her cabinet post.2 These anecdotes about cabinet appointments present an interesting contrast to the SNTV period. When SNTV was in place, the president chose to appoint his own followers inside the party, while not considering the optimal use of right talents with appropriate skills in the right places. For example, when President Chen Shui-bian reshuffled his cabinet in Jan- uary 2002, the media reported that, by assigning his own people and some political moderates to key posts, Chen eliminated all DPP leaders who had made contributions to his presidential election victory, to the extent that the newly appointed ministers lacked professional expertise (China Post, January 25, 2002, 19). The DPP's intraparty factional struggles went on and resulted in a series of party reform attempts to eliminate factionalism (Taipei Times, June 21, 2004, 4).3 But a crucial difference from Japan should be noted. Despite the epi- sodes above, which suggest that the Taiwanese party leadership can some- times strategically use cabinet positions for electoral purposes, it is still rare that former legislators are appointed to the cabinet, as figure 6.3 shows. Of the 503 ministers appointed since the late 1990s, only 11.3% had experi- ence serving in the legislature,4 suggesting that only on rare occasions were cabinet posts used as carrots for hard workers. Instead, as in other presi- dential democracies, cabinet posts usually went to technocrats, academics, and others from nonlegislative backgrounds, whereas in Japan it is rather rare to appoint nonlegislators to the cabinet.5 Party members using cabinet positions as stepping-stones for elec- toral positions are also still rare. For example, of the 68 cabinet members appointed by President Ma between 2008 and 2012, only five (7.4%) ran for election, and their performance was not impressive. Four of them ran in the 2012 Legislative Yuan elections: Yang Chih-liang and Yin Chi-ming were unsuccessful list-tier candidates, while Chiang Chi-chen and Su Jun- pin ran for SMDs in Taichung and Tainan, respectively, with only Chiang being successful. Chu Li-lun stepped down as deputy premier in 2010 to defeat the DPP's eventual 2012 presidential candidate, Tsai Ing-wen, in the New Taipei City mayoral election. 174 Mixed-Member Electoral Systems in Constitutional Context LO _ District Party list Other parties o LO 1997 2000 2001 2002 2005 2006 2007 2008 2009 2012 Cabinets Fig. 6.3. Proportion of former legislators as cabinet ministers in Taiwan, 1997-2009 (Note: Percentage is the number of incumbent district or party-list legislators divided by the total number of cabinet ministers. Because the premiers may appoint more than one person to a cabinet post, the denominators could be different from period to period. The KMT formed the cabinets between 1997 and 2000, and from 2008 to present.) Ultimately, these differences are the result of the constitutional dif- ferences between Japan and Taiwan. In contrast to parliamentary Japan, where cabinet appointments are much more closely aligned with a party's vote-seeking incentives, the president's survival in semipresidential Taiwan does not rely on legislative confidence, and therefore the incentive to stra- tegically use government positions to tame backbenchers is substantially weaker. Indeed, the Taiwanese president can appoint anyone to the cabi- net, including an opposition party member. For instance, President Chen appointed a retired air force general and KMT member Tang Fei as his first premier in 2000. In 2008, President Ma chose a former Taiwan Soli- darity Union (TSU) legislator, Lai Shin-yuan, as the chair of the Main- land Affairs Council, even though KMT backbenchers demanded that the president give priority to the party's loyal members when making cabinet appointments (Taipei Times, April 30, 2008, 5). In sum, the evidence we offer here suggests that parties in both Japan and Taiwan are now gradually making the strategic use of positions in the government for electoral purposes. This stands out in contrast with the SNTTV period, when the leadership had a much smaller voice in allocating posts because of the intraparty factional constraints. By promoting those Post Allocation, List Nominations, and Preelectoral Coalitions 17 5 who are loyal and demoting those who are disloyal, parties use government posts as carrots and sticks, although the decline of particularism caused by MA/FM seems to have made cabinet appointments less attractive as carrots. Nevertheless, the constitutional difference explains why Taiwanese presi- dents' use of these important governmental resources is still limited. List Nomination Strategy Candidate selection and nomination is also one of the party strategies to win more votes, as the nature of electoral systems systematically deter- mines what types of candidates are more popular. Therefore, we should observe how parties adjust their candidates accordingly (Hazan and Rahat 2010; Siavelis and Morgenstern 2008). As discussed by Yu, Shoji, and Batto in chapter 5, under Japan's SNTV incumbents were given endorsements as long as they wanted to run, as they were proven to be electorally powerful enough to win seats. When a member announced his or her retirement and an open seat was created, factional heads would engage in some bargain- ing but most often the retiring member's close relative (most often his or her son) was picked up to fill the vacancy for that faction (Ishibashi and Reed 1992). This is because koenkai were very costly to maintain and it was very difficult for the other factions to challenge those inherited seats from scratch. Thus, beyond giving nominations to incumbents or second- (or even third- or fourth-) generation hereditary candidates, parties' roles in the candidate recruitment and selection process were very minimal. The change to MMM should be expected to simplify this intraparty bargaining process among factions and therefore bring about some changes in parties' nomination strategies. One such change is the diminished influ- ence of factions. As Kobayashi and Tsukiyama show in chapter 3, now the leadership exercises more centralized power in nominating candidates in favor of its faction members. Another consequence is that the leadership can more strategically select those who would fit best into the new institu- tional environment. For instance, as Yu, Shoji, and Batto show in chapter 5, Japan's major parties now utilize the public recruitment (kobo) system to select attractive candidates on the nominal tier. In this section, we pri- marily focus on the list tier on which voters cast party votes. We expect that parties will try to nominate new candidates with fresh images, such as women or professionals who are well known nationwide, on the PR list in order to mobilize more party votes. In Japan, there are some signs that parties try to entice the voter with 176 Mixed-Member Electoral Systems in Constitutional Context attractive PR lists. When the LDP lost in the 1993 elections, it started discussing the introduction of an age limit for list candidates so that its PR lists would have a new image that could appeal to voters (Asahi Shimbun, September 8, 1993, 8). Although it took a couple of years for the leadership to convince veteran members, the party finally decided in 1995 to impose an age limit of 73, although when to actually impose it was made unclear (Asahi Shimbun, May 24, 1995, 2). After rounds of discussions, the LDP imposed this age limit beginning in the 2000 elections to all the candidates desiring to run on the list tier, except for the two former prime ministers- Yasuhiro Nakasone and Kiichi Miyazawa-who were guaranteed the top ranks in the list tier for life (Asahi Shimbun, January 28, 2000, 6). In 2003, these exceptions were finally dumped by Prime Minister Junichiro Koi- zumi, as the LDP needed to appeal to voters with "how the party was try- ing to transform itself' (Asahi Shimbun, October 15, 2003, 4). According to opinion polls, a great majority (77%) of voters supported this abolish- ment of the exceptions (Asahi Shimbun, November 2, 2003, 2). The DPJ followed the move in 2004, when it decided to impose the age limit at 70 to show the voters that it was more reform-minded than the LDP (Asahi Shimbun, February 16, 2004, 2). At the same time, the DPJ also decided to demote candidates' ranks in the list if they lost two consecutive elections in the nominal tier, while banning any relative within the third degree from inheriting a retiring member's district (Asahi Shimbun, May 2, 2004, 1),6 as this seat inheritance practice looks to the Japanese voter very undemocratic and unfair. Introducing the age limit was also designed to replace top-rank PR candidates with more attractive ones, as who sits atop a party's PR lists symbolizes whom the party is really meant to represent. In 2005, the LDP nominated 10 eye-catching female candidates with diverse career back- grounds at the top in most of the lists (Asahi Shimbun, August 30, 2005, 2). Described as "lipstick ninjas" armed with "blood-red lipstick and a killer smile" by the foreign media (Washington Post, September 3, 2005, 28), most of them were political amateurs without previous experience in the Diet. They were picked directly by Prime Minister Koizumi and his aides, so that they could give the voter the impression of change in the traditionally male-dominated party. Most of them (8 of 10) were also dually nominated at the nominal tier to compete with those who defected from the LDP to oppose Koizumi's postal reform agenda, giving the voter a choice of the old LDP vs. the new LDP.7 But some important institutional settings-dual candidacy and list rank- ing based on margins with the winner (sekihai ritsu)-somewhat constrain Post Allocation, List Nominations, and Preelectoral Coalitions 177 the leadership's list nomination strategies in Japan. The dual candidacy in Japan is unique in that candidates can be listed on the PR list at the same rank and their actual rankings are determined by how they perform in their SMDs, or, to be more precise, their vote margins with SMD winners. This dual candidacy at the same rank, or what is usually called the "best loser" provision, gives candidates the incentive to win as many votes as possible in their local SMDs, which is also compatible with parties' collective goal of seat maximization. But this provision makes it difficult to decide who should be prioritized in the list, as list-only candidates at safe ranks looked to SMD candidates as though they were free riders on the party (Asahi Shimbun, March 14, 2003, 4). For the same reason, it is also difficult to recruit high-profile professionals and celebrities as top-rank list candidates or to have a wide variety of list candidates recruited from different regions, different professions, and different genders, compared to Taiwan. Let us illustrate this problem with a typical party list shown in table 6.1. In this Tohoku PR bloc, the LDP won six seats. Four of the 28 can- didates were given favorable positions, while 23 candidates were given the same fifth rank. Note that eight candidates lost in their SMDs. However, of the eight losers, three were revived as zombies: Tokuichiro Tamazawa was favorably ranked in fourth place so that he got a sure win, even though he was unpopular in his district, having secured less than 40% of the votes relative to the invincible opposition leader, Ichiro Ozawa. Koji Futada and Atsushi Watanabe were also revived as they were close enough to the SMD winners compared to the other SMD losers from the LDP: for example, Masami Saito was also close, but unfortunately, his sekihai ritsu, or his votes relative to the winner, was lower than Futada's. This combination of a few favorably ranked candidates with a clump of same-rank candidates was a typical LDP strategy until the early 2000s. However, now this practice is increasingly rare, as it is difficult to convince other LDP candidates why, in the example shown in table 6.1, the four (Sakamoto, Nakano, Sato, and Tamazawa) should be prioritized over the rest. Now it is less common to give discrete ranks near the top of the list, where a party can guarantee a sure win to the candidates without their having to perform well in the nominal contest.8 As shown in figure 6.4, by 2005 and 2009, large parties' nomination and list-ranking strategies seemed to converge, as SMDs became more competitive. In Taiwan, the prohibition of dual candidacy on the single nationwide list tier with a district magnitude of 34 allows parties to more strategically use list candidates. The first strategy can be called a "regional reach-out strategy": parties can compensate for SMD outcomes with candidates from 178 Mixed-Member Electoral Systems in Constitutional Context certain regions, given the regional divide between the KMT-dominant north and the DPP-dominant south. Figure 6.5 shows that 48% of list legislators the DPP nominated were from the KMT's strongholds in the north: Taipei City, New Taipei City, Keelung City, Taoyuan County, Hsin- chu City and County, and Miaoli County, where the DPP expected uphill battles with the KMT. Although the DPP's party list in 2008 included only one former legislator from northern Taiwan (Ker Chien-ming), in 2012 its list included, in addition to Ker, Tuan Yi-kang, Wu Ping-jui, and Yu Tian, all of whom had been elected in the districts of northern Taiwan and therefore had some supporters in the region. Yu was the highest-ranked loser, while the other three all won list seats. In addition, Su Tseng-chang, the former premier, was listed at the marginal rank of 18, and party offi- TABLE 6.1. LDP's List for Tohoku Bloc, 2005, as a Typical Party List Dually nominated / Best loser's ratio Won Rank Name Inc Yes Yes Yes Yes Yes Yes Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) Yes (in SMD) 1 2 3 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 28 Goji Sakamoto Masashi Nakano Tatsuo Sato Tokuichiro Tamazawa Atsushi Watanabe Koji Futada Akinori Eto Tadamori Oshima Taro Kimura Shun'ichi Suzuki Toru Doi Ken'ya Akiba Akihiro Nishimura Shintaro Ito Itsunori Onodera Nobuhide Minorikawa Toshiaki Endo Takehiko Endo Koichi Kato Yoshitami Kameoka Takumi Nemoto Masayoshi Yoshino Masami Saito Atsushi Oikawa Hidenori Hashimoto Takaki Ono Susumu Hasumi Takao Sato cumbent Gender Yes Male Yes Male Yes Male Yes Male Male Yes Male Yes Male Yes Male Yes Male Yes Male Male Yes Male Yes Male Yes Male Yes Male Yes Male Yes Male Yes Male Yes Male Male Yes Male Yes Male Male Male Male Male Yes Male Male (where applicable) Yes / 38.6% Yes / 92.7% Yes / 88.8% Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes / 87.6% Yes/68.5% Yes/68.1% Yes/66.1% Yes / 45.9% Post Allocation, List Nominations, and Preelectoral Coalitions 179 20.0% 18.0% 16.0% 14.0% 12.0% ... 10.0% . LDP 8.0% DPJ 6.0% 4.0% ' . 2.0% 0.0% 1996 2000 2003 2005 2009 2012 Fig. 6.4. Favorably ranked candidates from the LDP and DPJ ials said that the very influential figure inside the party was strategically placed just outside the safe zone because he was expected to mobilize his supporters in New Taipei City, where he had previously headed the local government (Taipei Times, June 30, 2011). To further explore how (dis)proportionally the two major parties nomi- nate regional candidates on the list tier, we group the KMT's and DPP's list members in 2008 and 2012 by six regions and then put them in a scat- ter plot shown in figure 6.6. It is apparent that both parties did not recruit their list members in proportion to the population, with the exception of Taipei City and its suburban area. The DPP drew six list members from the least populated area-Yilan, Hwalien, and Taitung-while the KMT found 12 list members in Kaohsiung, Pingtung and Penghu, part of south- ern Taiwan. Compared to figure 6.5, which counts all of the list members after 1992, the result implies that the KMT has targeted southern Taiwan since the electoral reform. In contrast to this regional strategy, a party can also choose what can be called a "group reach-out strategy": it can step up efforts to attract votes from a wide range of social segments by nominating a variety of represen- tatives from disadvantaged and special interest groups. Compared to 2008, when the KMT's list candidates were mostly those incumbents who had to find places to run because of the downsizing of the Legislative Yuan (Taipei Times, November 8, 2007), the party in 2012 ambitiously tried to reach out to different minority groups. Seeing how the DPP reserved the second 180 Mixed-Member Electoral Systems in Constitutional Context 100- 75 - KMT stronghold 0 50- DPP stronghold Special groups 25- 0- KMT DPP Political Parties Fig. 6.5. Percentages of KMT's and DPP's list members from their own strongholds, their rival's strongholds, or special groups, 1992-2012 (Note: List members who have run in elections or served in the party organizations in the city/ county, or have been associated with the place, are considered influential in their own locale. The KMT's strongholds are defined as the city/county where the DPP's vote share is lower than the DPP's average national vote share. The DPP's strongholds are defined as the city/county where the DPP receives a higher vote share than its national one. Special groups refer to mii- tary, teachers, judges and social groups. N = 157.) spot for Chen Chieh-ju, president of the Yu-Cheng Social Welfare Foun- dation, and the third spot for Wu Yi-chen, a women's rights lawyer, the KMT reserved the second, third, and fourth spots for Wang Yu-min (the executive officer of the Child Welfare League Foundation), Tseng Chu-wei (a well-known public finance scholar), and Yang Yu-hsin (the secretary- general of the Taiwan Foundation for Disadvantaged Patients), respec- tively. The media and even the DPP's presidential candidate Tsai Ing-wen Post Allocation, List Nominations, and Preelectoral Coalitions 181 a-) E - 0 c - (0l- South (Kaohsiung City, Kaohsiung County, ' Pingtung County and Penghu County) North (Taipei Cityrtaipei County and,Keelung City) South-Central (Yunlin County, Chiayi City, Chiayi County, Tainan City and Tainan Cdunty) North-Central (Taoyuan County, Miaoli County. Hsinchu Qity and Hsinchu County) /Central (Taichung City, Taichung County, Changhua County and Nantou City) East (Yilan County. Hwalien County and Taitung County) 0 0 2 4 6 8 10 12 Number of DPP's List Members Fig. 6.6. Number of two parties' ist members dustered by regions, 2008 and 2012 (Note: Circles are in proportion to the population of each region, 2010.) highly praised this aggressive recruitment: for example, the China Post (November 18, 2011) reported Tsai as saying: "It is good to see the [KMT] come up with a more diverse list of candidates for the at-large legislative seats, including activists representing underprivileged groups. . . . We do hope that the KMT nominations are more than just an election gimmick. Let's hope that it really marks a change of the party's mentality."9 Note that these ambitious list nomination strategies cannot be expected to occur in Japan now, where major parties simply dually nominate on their lists SMD candidates, most of whom are from their local districts. One clear illustration of this is how parties treat women in Japan and Taiwan. As most of the list candidates are dually nominated now and male candidates are still dominant in the nominal tier, significantly fewer female candidates are nominated in Japan than in Taiwan: between 1996 and 2012, females accounted for only 5.1% and 11.3% of the LDP's and the DPJ's list can- 182 Mixed-Member Electoral Systems in Constitutional Context didates, respectively. Meanwhile in Taiwan, even before the Constitution was amended in 2005 to require that half of list legislators must be women, political parties attempted to increase the share of female party-list legisla- tors to improve their image. In the 2004 election for instance, the DPP allocated six out of 16 seats to women.'0 In sum, with the transition to MAMM, parties now have strategic reasons to offer attractive sets of candidates in the list tier. In light of this, Japanese parties started adopting age limits and installing fresh-image candidates and widely known professionals at the top. But with the best loser provi- sion, Japan's parties now just choose to use the list tier only as a safety net for SMD candidates, by dually nominating almost all of them simply at the same rank so that they can maximize votes and seats on the nominal tier. This has resulted in the declining number of favorably ranked candidates. In contrast, parties in Taiwan can strategically nominate females, nationally well-known figures, and even candidates from the main opponent's strong- holds to boost votes across the country. Candidate Coordination Strategy The third among various party strategies for vote mobilization that we discuss in this chapter is about candidate coordination, or preelectoral coalitions. Under SNTV, it is usually difficult for preelectoral coalitions to emerge for two reasons. First, if M is large enough-say, five-then all a party needs to do to maintain one seat is to secure a small fraction of the votes (16.7%), so even a small party winning only 10-15% of the votes has a good chance to be competitive and will not easily give up running candidates. Second, intraparty factionalism accentuated by SNTV nec- essarily increases the interparty bargaining complexities. This is because even if a leader chooses to withdraw a candidate from a district, the leader needs to convince one of the factional bosses to give up running. Thus, we expect that, with the transition to MMM, the mechanical and psychologi- cal effects of Duverger's Law and the simplified intraparty bargaining pro- cess for candidate nomination should help party leaders to directly bargain with the other parties over candidate entry. Large parties take advantage of the majoritarian nature of MMM to nominate candidates in almost all the SMDs while convincing smaller parties to join preelectoral coalitions." Therefore the effective number of electoral parties in each district should converge to two over time. This is our general expectation. However, there are several theoretical caveats to this general expecta- Post Allocation, List Nominations, and Preelectoral Coalitions 183 tion. One is about the presence of the list tier. If joint lists are not allowed, then preelectoral coalition partners would need to compete on the list tier. LDP candidates, for example, usually ask their own personal supporters to cast their list votes for the party's coalition partner, the Clean Government Party (Komeito, CGP), to make sure that CGP supporters would vote for them in exchange (Asahi Shimbun, September 7, 2000, 4). However, this is nothing more than lip service. Because joint lists are not allowed in Japan, unlike in Taiwan, there is really no way to go about ordering voters to split their tickets at the booths: the exit polls in 2003, for example, suggest that only 6% of LDP supporters voted for the CGP on the list tier (Asahi Shim- bun, November 10, 2003, 3). One former CGP legislators lamented: "The coalition should be 'give and take' but we only give in to the LDP" (Nihon Keizai Shimbun, August 27, 2005, 2). This is consistent with the empirical evidence presented in chapter 7 by Wang, Lin, and Hsiao. In addition to this fear of being exploited by large parties, minor parties want to improve their performance on the list tier by using hopeless SMD candidates and "contaminating" the list tier (Herron and Nishikawa 2001). In Japan, the Communists nominated hopeless SMD candidates in almost all the SMDs except in 2009, even though it could win only a couple of SMD seats back in 1996. In addition to this extreme case, the Social Democratic Party (SDP) in 2009 also had to compete with the DPJ in nearly half of the 31 SMDs where it ran candidates.'2 This limited coordination is partially the result of the SDP's reliance on the list tier for its survival: "The SDP was agonized over how to distance itself from its larger partner, because it might end up encouraging more voters to cast PR votes for the DPJ by supporting the DPJ's SMD candidates" (Asahi Shimbun, August 30, 2009, 2). Thus, in a mixed-member system, preelectoral coalition partners face a commitment problem: smaller parties fear that their larger partners might exploit votes, while larger parties cannot be sure if their smaller partners' activists will be really collaborative in local districts. Of course, this com- mitment problem should be omnipresent under any single-member sys- tem, but the presence of the list tier, on which small parties rely for their electoral survival, somewhat aggravates the problem. Thus, there must be some credible commitment mechanisms, and we argue that Japan's insti- tutional arrangements-parliamentarism and bicameralism-somehow helped to solve, if not perfectly, the commitment problem. That is why we see the gradual convergence of the effective number of electoral parties to two, as shown in figure 6.7, although the case of the 2012 elections looks very anomalous. As shown by Lin in chapter 2, parliamentarism can help coalition part- 184 Mixed-Member Electoral Systems in Constitutional Context C0 1996 2000 2003 2005 2009 2012 Fig. 6.7. Effective nurnber of electoral parties in each district in Japan, 1996- 2012 ners become united as a single team by taking mutual hostages. Small par- ties can offer support for the confidence vote and the government's survival on the one hand, while they can influence policy making and implemen- tainthrough positions inthe government onthe other. Thus, in ever cabinet after Obuchi's second reshuffle in 1999, the LDP appointed one minister and several vice ministers from the CGP, and when it did so, it incorporated the CGP's demands for positions (Asahi Shimbruni, Septem- ber 28, 2004, 3). Another solution is to forge detailed policy agreements before the election to make sure everyone signing the agreement will join the government after a turnover. Before the 2009 elections, the DPJ had rounds of policy negotiations with two potential coalition partners (the SDP and the People's N\ew Party [PNP]) to develop common policy plat- forms (Nihon Keizai Shimbin, May 28, 2009, 2). In an interview before the 2012 elections, the CGP's leader, Natsuo Yamaguchi, emphasized that, although his party and the LDP might have had some policy differences, "the two parties share the history of forming coalition governments. . . . So we can reach policy agreements that the people can rely on" (Asahi Shim- bun, November 30, 2012, 4). Bicameralism in Japan also helped to make these coalitions more stable. Post Allocation, List Nominations, and Preelectoral Coalitions 185 Japan's upper house, the House of Councillors (HC), has equal lawmak- ing power with the lower house except over the budget and the investi- ture vote. The HC features a more proportional and therefore fragmented MA/FM, in which the list tier uses a nationwide open-list PR district with M around 50. No party has won a single-party majority in the HC since 1989. Thus, even when the LDP retained a single-party majority in the lower house from 1997 to 2009, it always needed legislative support from the CGP. Similarly, the SDP and the PNP had a considerable bargaining advantage with their 11 seats in the HC. Therefore, except in 2012, small parties strategically coordinated can- didates where possible with two of the major parties (the LDP or the DPJ), resulting in the Duvergerian convergence shown in figure 6.7. The LDP and the CGP first formed a coalition government in 2000, and this coali- tion spread to district-level cooperation. Although the two coalition part- ners competed against each other in some of the districts in the 2000 elec- tions, they stopped doing so after 2003 in any of the 300 SMDs. Where the LDP is traditionally strong, the CGP refrained from running candidates, and the LDP's SMD candidates can expect votes from a large majority of the CGP supporters (Asahi Shimbun, June 26, 2000, 3). Meanwhile, the LDP avoids sending candidates to the CGP's strongholds, where the CGP's loyal supporters, based on a religious group Soka Gakkai, are geographi- cally concentrated.3 Beginning with the 2007 upper-house elections, the DPJ also coordinated candidates with some minor parties (the SDP and the PNP) to maximize the possibility of a government turnover through elections. Unlike the LDP-CGP coalition, this opposition coalition could not perfectly coordinate candidates as described above, but it was still suc- cessful in gradually reducing the number of contested districts from 41 (13.7% of 300 SMDs) in 2003 to 30 (10.0%) in 2005, and then down to 14 (4.7%) in 2009. Where the DPJ decided not to run in 2009, it officially backed the smaller partner's candidates (Asahi Shimbun, July 22, 2009, 4). It is still yet to be seen whether the fragmentation in 2012 will continue or not. In a sense, this is something one should expect under a parliamen- tary regime: as Batto and Cox's introductory chapter in this volume argues, a party might choose to independently run as a third party and try to join the cabinet by engaging in postelectoral bargaining with one of the major parties if no party wins a majority. In line with this view, the results of 2012 were mainly caused by the rise of minor parties, the so-called third forces (daisankyoku) seeking votes by disassociating themselves from the exist- ing parties and attracting nonpartisan voters disillusioned with the very unsatisfactory performance of the three-year DPJ administration. One of 186 Mixed-Member Electoral Systems in Constitutional Context them, the Japan Restoration Party (Nippon Ishin no Kai, JRP), was quite successful with 54 seats (11.3 %), while others were moderately successful: The Your Party (Minna no To) and the Tomorrow Party of Japan (Nippon Mirai no To) won 18 (3.8%) and 9 seats (1.9%), respectively. Many of the SMDs were marked by three- and even four-way competition, disturbing the trend toward the Duvergerian equilibrium as shown in figure 6.7. But we would also like to note that, although the parliamentary post- election bargaining story as posed by Batto and Cox may explain well the rise of third parties in 2012, it has some difficulties in explaining the grad- ual decline in the number of parties until 2009. In fact, most of the third parties, except the ideologically extreme Communists, were merged into one of the two major camps led by the LDP and the DPJ by 2003.14 And, even in 2012, the minor parties themselves stepped up their efforts to form the united front against the LDP-CGP coalition: in fact, the JRP and the Tomorrow Party of Japan were born when even smaller parties decided to merge," while the JRP and Your Party tried to coordinate their candi- dates in as many districts as possible until the last minute (Asahi Shimbun, December 1, 2012, 4). So it might be the case that the 2012 elections were just a temporal outlier. Several of the minor parties might be merged or join either of the two large camps before too long, but as of this writing it is still very unclear. Let us move to the case of the MMM system in semipresidential Tai- wan, which similarly indicates that in most of the districts, Duverger's Law is working: the average effective number of candidates stood at 2.09 and 2.18 in 2008 and 2012, respectively. However, this masks how preelectoral coalitions work in reality. In 2008, the small parties in the Pan-Blue coali- tion refrained from nominating SMD candidates, but the TSU competed with the DPP in 13 districts (17.8% of the 73 districts) although no candi- date from the TSU could win. In 2012, it was the opposite: The Pan-Blue coalition was split, as the People's First Party nominated SMD candidates in 10 districts (13.7%) to challenge the KMT, and no candidate could win seats, just like the TSU in 2008. This stands out in stark contrast to Japan, where, at least until 2009, the two large camps were successful in coordi- nating candidates so that less than 5% of the districts observed intracamp competition. This could be because of the heavy regulations on which parties can run for the list tier,16 but we argue that this is because of the presence of (nega- tive) presidential coattails: as the presidency is the most important political prize in presidential and semipresidential democracies, the media and the general public pay the greatest attention to the presidential race and presi- Post Allocation, List Nominations, and Preelectoral Coalitions 187 dential candidates' performance significantly affects legislative elections as well (Golder 2006). Without viable presidential candidates to start with, the small parties have a very strong incentive to associate themselves with a major party in the same coalition when it has a very winnable candidate. This scenario happened in 2008, when the KMT's Ma Ying-jeou looked like a sure winner well before the 2008 elections (Chu 2008; Tan 2009). The People First Party (PFP) readily formed a preelectoral coalition with the KMT and had a joint PR list-four PFP members were placed in the KMT's party list and three of them were ranked high enough to win seats. At the same time, the PFP was very collaborative on the nominal tier as well, as it did not run candidates in any SMDs (Chu 2008, 126). Similarly in 2012, the TSU was very collaborative with the DPP: learning from 2008, the TSU this time decided not to run on the nominal tier and instead yield to its Pan-Green alliance partner (Taipei Times, May 29, 2011). The TSU's spiritual leader, Lee Teng-hui, even called on voters to cast their ballots for the DPP's SMD candidates and give their party votes to the TSU (Taipei Times, December 26, 2011). This was because a series of opinion polls sug- gested that the DPP's presidential candidate, Tsai Ing-wen, was running neck and neck with Ma. As late as December 10, 2011, about a month before the election, a TVB super channel poll showed Tsai and Ma tied with 39% each. Another poll by a Taiwan think tank showed Tsai leading by 3 5.9% to 3 5.4% (Taipei Times, December 10, 2011). Riding the coattails of the popular presidential candidate from the same camp, the TSU won three PR seats, a breakthrough from its zero seats in the previous election. On the other hand, small parties similarly have a very strong incentive to dissociate themselves from the major party when the coalition's pres- idential candidate is unpopular, and this happened to the TSU in 2008 and the PFP in 2012. Before the 2008 election, Chen Shui-bian was very unpopular as president and leader of the DPP, with the ever-worsening economy, increasing cross-strait tensions with mainland China, and a series of corruption scandals. Even the DPP leadership predicted devastating defeats in the 2008 elections (Chu 2008; Tan 2009). This created a schism in the Pan-Green coalition, with Lee Teng-hui explicitly criticizing the DPP as "incompetent, preoccupied with winning the presidential election" (Taipei Times, February 14, 2007). In an interview, Lee went on to charac- terize President Chen as having "no concept of democracy" and worried that Taiwan was being held hostage politically by the two main political parties (Taipei Times, December 29, 2007). As the TSU tried to dissociate itself from the unpopular president, TSU chairman Huang Kun-huei did not like the idea of withholding its candidates on behalf of the DPP (Taipei 188 Mixed-Member Electoral Systems in Constitutional Context Times, July 20, 2007). The schism in the Pan-Green coalition was never resolved despite the common belief that "the Pan-Green Coalition will lose the elections if there is no cooperation with each other" (Taipei Times, August 4, 2007). DPP chairman Yu Shyi-kun often attempted to convince the smaller partner to "only nominate the most promising candidates for the next legislative election based upon public opinion survey results," describing the two parties as "brothers in a family" (China Post, January 31, 2007), but to no avail. After its landslide victory in 2008, the KMT faced a similar problem. One of the key events was Typhoon Morakot in August 2009, which sig- nificantly undermined the perception of the administration's ability to pre- pare for a natural disaster and Ma's credibility as a leader (Gold 2010). As Ma's popularity steadily decreased over time (Taipei Times, April 21, 2011), the KMT began losing votes and seats. For example, the mayoral elections in November 2010 showed that its vote share decreased by 7.6% compared to that of 2005-6. Even though the KMT maintained some major city mayoral positions, this poor performance was an ominous bellwether for the coming presidential race (Tien and Tung 2011). Sensing the worsening situation around the Pan-Blue coalition, PFP chairman James Soong said that his party would nominate at least 10 candidates for the legislative elec- tions as early as June 2011 (Taipei Times, June 11, 2011). Time after time the KMT tried to convince the PFP to collaborate on candidate coordina- tion (Taipei Times, July 27, 2011) and even suggested PFP candidates could run where the KMT had not yet completed its nomination process (Taipei Times, July 30, 2011). But ultimately the PFP decided to run its own candi- dates for presidential, SMD, and party-list elections. In sum, with the transition to MMM, in theory we should observe how small parties will try to form preelectoral coalitions in the nominal tier in order to survive. But coordination cannot go smoothly unless there is a commitment mechanism, because small parties fear exploitation by larger partners and want to maximize list votes with the contamination effects. Thus, Japan's larger parties (the LDP and the DPJ) offered cabinet posi- tions to and forged coalition agreements with small parties (the CGP, the SDP, and the PNP) to make preelectoral coalitions stable. Consequently, the number of candidates at the district level gradually declined at least until 2009. Lacking such mechanisms, semipresidential Taiwan, by con- trast, was marked by limited coordination. Rather, small parties (the TSU in 2008 and the PFP in 2012) challenged their larger partners when they sensed that their odds of winning improved by distancing themselves from their camps' presidential candidates. Post Allocation, List Nominations, and Preelectoral Coalitions 189 Conclusion: Constitution, Dual Candidacy, and Beyond In this chapter we investigated how parties have strategically adapted them- selves to the new electoral environment in Japan and Taiwan by focusing on three different issues: post allocation, list nomination, and candidate coordination. The main argument is that the simplified intraparty bargain- ing process over posts and candidates under MMM allows the leadership to use available tools and resources more strategically for electoral pur- poses than it could during the SNTV period. More specifically, in terms of post allocation, ruling parties should make more strategic use of positions in the government as carrots and sticks to control candidates. In terms of nominations, parties should strategically nominate attractive candidates on the list tier to boost party votes. And in terms of candidate coordination, parties should strategically form preelectoral coalitions to minimize wasted votes. In general, parties in both Japan and Taiwan are moving toward these expectations. However, throughout this chapter we also find the institutional differ- ences between the two countries-parliamentarism versus semipresiden- tialism and the dual candidacy provision-are so significant that they can outweigh these naive predictions driven by the theory of electoral systems alone. First, mutual dependence between the cabinet and the legislature for survival in parliamentary Japan results in a much closer alignment between the vote-seeking incentives and cabinet appointment patterns: now the prime minister appoints fewer and fewer zombies to important positions in the government so that candidates can win more SMDs. But in Taiwan, although some signs suggest that the president rewards hard workers with cabinet positions to develop quality candidates in the opposition's strong- holds, they are still rare, because constitutionally the president really does not depend on the party or the legislature for support. Second, to maximize the strategic use of the best loser provision, Japan's parties now use the list tier only as a safety net for SMD candidates. This is because the leadership can expect to maximize its nominal-tier winners by dually nominating SMD candidates at the same top rank and letting them engage in laissez-faire competition on a level playing field. Now par- ties nominate very few favorably ranked candidates. In Taiwan, the single nationwide list tier without dual candidacy allows for more strategic nomi- nating and list-ranking of candidates. Therefore party lists in Taiwan tend to include a wider variety of candidates from different regions, genders, and career backgrounds than in Japan. Third, parliamentarism can effectively address the credible commit- 190 Mixed-Member Electoral Systems in Constitutional Context ment problem that could otherwise make candidate coordination difficult. Japan's larger parties (the LDP and the DPJ) offered cabinet positions to and forged coalition agreements with small parties (the CGP, the SDP, and the PNP) to make preelectoral coalitions stable. Therefore Japan observed the gradual convergence to the Duvergerian equilibrium as evidenced by the declining number of candidates at least until 2009. Lacking such mechanisms, semipresidential Taiwan was, by contrast, marked by limited coordination. Rather, small parties (the TSU in 2008 and the PFP in 2012) challenged their larger partners when they did not like their camps' presi- dential candidates. All of these ideas imply that the effects of electoral rules are mediated by contextual factors. It is important to look beyond district magnitude, ballot types, and seat allocation methods when understanding parties' vote mobilization strategies. Although this chapter's primary focus is on the two main institutional differences between Japan and Taiwan, one could point to other contextual differences, including but not limited to, ethnic and linguistic heterogeneity, the degrees of democratic consolidation, and the presence of a second chamber (cf. Moser and Scheiner 2012). While we did not delve into how these different contextual factors could affect par- ties' mobilization strategies, we believe these will open up new research opportunities for comparing Japan, Taiwan, and any other mixed-member systems from different perspectives. NOTES 1. To this ladder one could add vice chair and chair of a Policy Affairs Research Council subdivision between (II) and (III), as the literature does (Ramseyer and Rosenbluth 1993, 86), but we do not do this since the Policy Affairs Research Council became less and less important, especially after Koizumi took power in 2001. In addition, since Japan has no clear notion of demotion, unlike in other parliamentary democracies, we focus on only promotion here. 2. The KMT finally nominated a party-list legislator from Tainan, Chen Su- hui, to run in the by-election against the DPP nominee, former mayor of Tainan City, Hsu Tian-Tsair. Hsu defeated Chen by a margin of 61% to 38%. 3. Also see chapter 4 by Batto and Huang in this volume for a much more detailed account of the DPP's factional struggles, especially between the New Tide faction and the other factions over cabinet positions, and its attempts to reform itself. 4. This may sound a bit contradictory with Lin's evidence in chapter 2, but here we look at whether ministers have served as legislators before. 5. On average, the Japanese cabinet has only one or even no nonlegislator, with some exceptions including Koizumi's first cabinet (2001-2), which had three non- Post Allocation, List Nominations, and Preelectoral Coalitions 191 legislators (16.7%) in 18 cabinet positions, and Fukuda's first cabinet (2007-8), which had two. 6. This means that a retiring incumbent cannot designate as a successor can- didate his/her child (first degree), sibling or grandchild (second degree), or niece/ nephew or great-grandchild (third degree). In order to differentiate itself from the LDP, the DPJ in 2012 applied this strict rule to any candidate hoping to run in SMDs, including Yuichiro Hata, who showed an interest in succeeding his retiring father, former prime minister Tsutomu Hata. Denied an endorsement from the party, the younger Hata chose not to run (Nihon Keizai Shimbun, November 18, 2012, 2). 7. Of the eight dually nominated candidates, three won their SMDs, while the other five were elected on the PR list thanks to their high rankings. 8. Favorably ranking some candidates near the top of the list is more of a legacy of the old SNTV system (see Reed 1995). As the same district had more than one incumbent under SNTV and the leadership needed to deal with them under the new system, those who could not win nominations for the SMD race were given favorable ranks in the list. 9. There are two main reasons why the KMT's list was praised more than the DPP's list. The first reason is that many of the younger nominees on the DPP list had been either former cabinet ministers or legislators, while the KMT recruited many new faces from different fields and areas. Although Tsai argued that those former legislators or ministers could more effectively represent the interests of minority groups, it was a tough sell for the mass media. The second reason is that the KMT's list was completely different from the conservative-looking 2008 list. In that list, the KMT filled the first 20 spots with 17 incumbent legislators, two cabinet ministers, and one businessperson. 10. Prior to the 2005 constitutional reform, any party winning at least five seats was required to reserve at least one seat for a woman. 11. This also means that under a much more proportional MMP system, where list votes determine parties' seat shares in the parliament, preelectoral coalitions at the district level are not likely to occur. On this point, see chapter 9 by Shugart and Tan in this volume on New Zealand's MMP. 12. In these 15 districts where the DPJ and the SDP competed, the DPJ won 14 seats. In one district where the LDP won, the DPJ could not win even if hypotheti- cally the SDP gave up running a candidate. 13. Except in 2009, when none of the CPG SMD candidates survived, the party has been remarkably successful in SMDs for a relatively small party: in 45 districts where the CPG ran candidates between 2000 and 2012, it won 33 seats (73.3%). This is because the party has been very good at predicting the size of its supporter bases and optimizing candidate entry accordingly, using its religious-group net- work (Lee 1970, 507). 14. They were mainly the remnants of the second largest party in the 1996 elec- tion, the New Frontier Party, which was suddenly dissolved in 1997 into several minor ones: the New Fraternity Party (Shinto Yuai), the Good Governance Party (Minsei To), the Liberal Party (Jiyu To), and others joined the DPJ, while the Con- servative Party (Hoshu To) and others chose the LDP. 15. The JRP was born in September 2012 when the populist Osaka mayor, 192 Mixed-Member Electoral Systems in Constitutional Context Toru Hashimoto, extended his Osaka-based regional party, the Osaka Restoration Association, into a nationwide one. In addition to about a dozen MPs individually switching to the JRP, some other tiny parties tried to join the force, including the Tax Cut Japan (Genzei Nippon) and the Sunrise Party (Tachiagare Nippon), and the latter finally merged in November after rounds of negotiations. The Tomor- row Party of Japan was born in late November when several DPJ splinters-the Kizuna Party (Shinto Kiduna), People's Life First (Kokumin no Seikatsu ga Dai- ichi), Green Wind (Midori no Kaze), and Tax Cut Japan, which was rejected by the JRP-merged into a single party. 16. More specifically, a party can run its list if any of the following four conditions hold: (1) its presidential candidate got at least 2% of the votes in the most recent election; (2) it won at least 2% of the list votes in the previous three elections; (3) it has at least five legislators; or (4) it nominates at least 10 SMD candidates. REFERENCES Akarui Senkyo Suishin Kyokai. 2010. Dai 45-kai Shugiin Giin Sosenkyo noJittai [The results of the 45th House of Representatives election]. Tokyo: Akarui Senkyo Suishin Kyokai. Chu, Yun-han. 2008. "Taiwan in 2007: The Waiting Game." Asian Survey 48 (1): 124-32. Gold, Thomas B. 2010. "Taiwan in 2009: Eroding Landslide." Asian Survey 50 (1): 65-75. Golder, Matt. 2006. "Presidential Coattails and Legislative Fragmentation." Ameri- can Journal of Political Science 50 (1): 34-48. Grofman, Bernard. 1999. "SNTV: An Inventory of Theoretically Derived Proposi- tions and a Brief Review of the Evidence from Japan, Korea, Taiwan, and Ala- bama." In Elections in Japan, Korea, and Taiwan under the Single Non-Transferable Vote: The Comparative Study of an Embedded Institution, ed. Bernard Grofman, Brian Woodall, Sung-Chull Lee, and Edwin Winckler, 375-416. Ann Arbor: University of Michigan Press. Hazan, Reuven Y., and Gideon Rahat. 2010. Democracy within Parties: Candidate Selection Methods and Their Political Consequences. Oxford: Oxford University Press. Herron, Erik S., and Misa Nishikawa. 2001. "Contamination Effects and the Num- ber of Parties in Mixed-Superposition Electoral Systems." Electoral Studies 20 (1): 63-86. Ishibashi, Michihiro, and Steven R. Reed. 1992. "Second-Generation Diet Mem- bers and Democracy in Japan: Hereditary Seats." Asian Survey 32 (4): 366-79. Lee, Jooinn. 1970. "Komeito: Sokagakkai-Ism in Japanese Politics." Asian Survey 10 (6): 501-18. McElwain, Kenneth Mori. 2012. "The Nationalization of Japanese Elections." Journal of East Asian Studies 12 (3): 323-50. Moser, Robert G., and Ethan Scheiner. 2012. Electoral Systems and Political Context: How the Effects of Rules Vary across New and Established Democracies. Cambridge: Cambridge University Press. Nemoto, Kuniaki, Ellis S. Krauss, and Robert Pekkanen. 2008. "Policy Dissension Post Allocation, List Nominations, and Preelectoral Coalitions 193 and Party Discipline: The July 2005 Vote on Postal Privatization in Japan." British Journal of Political Science 38 (3): 499-52 5. Noble, Gregory W. 2010. "The Decline of Particularism in Japanese Politics." Journal of East Asian Studies 10 (2): 2 39-7 3. Patterson, Dennis, and Ko Maeda. 2007. "Prime Ministerial Popularity and the Changing Electoral Fortunes of Japan's Liberal Democratic Party." Asian Sur- vey 47 (3): 415-33. Pekkanen, Robert, Benjamin Nyblade, and Ellis S. Krauss. 2006. "Electoral Incen- tives in Mixed-Member Systems: Party, Posts, and Zombie Politicians in Japan." American Political Science Review 100 (2): 183-93. Ramseyer, J. Mark, and Frances McCall Rosenbluth. 1993. ]apan's Political Market- place. Cambridge: Harvard University Press. Reed, Steven R. 1995. "The Nomination Process for Japan's General Election: Waiting for the Heiritsu-sei." Asian Survey 35 (12): 1075-86. Reed, Steven R., Ethan Scheiner, and Michael E Thies. 2012. "The End of LDP Dominance and the Rise of Party-Oriented Politics in Japan." journal ofapa- nese Studies 38 (2): 3 53-76. Rochon, Thomas R. 1981. "Electoral Systems and Voting Behavior: The Case of Japan." In Parties, Candidates, and Voters in Japan, ed. John C. Campbell, 1-28. Ann Arbor: Center for Japanese Studies, the University of Michigan. Rosenbluth, Frances, and Michael Thies. 2010. Japan Transformed: Political Change and Economic Restructuring. Princeton: Princeton University Press. Shugart, Matthew Soberg. 2001. "Electoral 'Efficiency' and the Move to Mixed- Member Systems." Electoral Studies 20 (2): 173-93. Siavelis, Peter M., and Scott Morgenstern, eds. 2008. Pathways to Power: Political Recruitment and Candidate Selection in Latin America. University Park: Pennsyl- vania State University Press. Tan, Alexander C. 2009. "The 2008 Taiwan Elections: Forward to the Past?" Elec- toral Studies 28 (3): 502-6. Tien, Hung-Mao, and Chen-Yuan Tung. 2011. "Taiwan in 2010." Asian Survey 51 (1): 76-84. SEVEN Split-Ticket Voting under MMM T Y Wang, Chang-chih Lin, and Yi-ching Hsiao Split-ticket voting has generated substantial interest in the discipline as demonstrated by the voluminous publication on the subject (Alesina and Rosenthal 1995; Bawn 1993, 1999; Beck et al. 1992; Born 1994; Brandy 1993; Cox 1990, 1997; Cox and Schoppa 2002; Duverger 1959; Feigert 1979; Ferrara, Herron, and Nishikawa 2005; Fiorina 1996; Fisher 1973; Gallagher 1998; Garand and Lichtl 2000; Gschwend, Johnston, and Pat- tie 2003; Herron and Nishikawa 2001; Huang 2001; Huang, Chen, and Chou 2008; Jacobson 1991; Jesse 1988; Karp et al. 2002; Kohno 1997; Lewis-Beck and Nadeau 2004; Reed 1999; Roscoe 2003). With a focus on electoral behavior in the United States where elections on executive and legislative bodies are held simultaneously, ticket-splitting is generally defined as a voting behavior where voters cast two or more than two bal- lots in an election to candidates of different parties competing for differ- ent positions. For instance, when elections for the presidency and for the House of Representatives in the United States take place at the same time, voters may choose to vote in the manner of a straight ticket by casting both ballots for candidates from the same party, or to split the ticket by voting for candidates from different parties on different ballots. Split-ticket voting can also occur in countries with mixed-member sys- tems for legislative elections (Choi 2006; Gschwend, Johnston, and Pattie 2003; Huang 2008a, 2008b; Huang, Wang, and Kuo 2008; Karp et al. 2002; Kohno 1997; Moser and Scheiner 2004, 2009; Reed 1999; Schoen 1999). 194 Split-Ticket Voting under MMM 195 Characterized as hybrids that combine both proportional representation (PR) and single-member district (SMD), mixed systems have been adopted in many countries including Germany, Hungary, Japan, New Zealand, and more recently, Taiwan. Depending on whether there is a linkage between votes acquired from the PR and from the SMD tier, the institutional designs can be further categorized as "nonparallel" mixed-member pro- portional (MMP) and "parallel" mixed-member majoritarian (MMM) sys- tems (Bawn 1993, 1999; Cox 1997, 81-83; Cox and Schoppa 2002; Fisher 1973; Gschwend, Johnston, and Pattie 2003; Karp et al. 2002; Shugart and Wattenberg 2001, 13-14). In nonparallel MMP systems, as in Germany and New Zealand, votes acquired from the PR tier are used to balance the overall representation of parties in the legislature so that weaker parties receive a number of seats proportionate to their general performance in the election. The majoritarian boost received by a large party from the SMD tier is likely to be reduced by the proportional allocation from the party list. Because the PR votes are used to compensate for the SMD votes, the nonparallel MMP is also characterized as "compensatory" mixed sys- tem (Karp et al. 2002; see also chapter 9 by Shugart and Tan, this volume). In MMM systems, as in Japan and Taiwan, seats in the SMD and PR are allocated independently and separately with no linkage between the two tiers. They are thus characterized as "parallel" systems. Despite these dif- ferences, both types of mixed systems feature electoral rules that provide voters with two ballots. One ballot determines which candidate wins seats in SMD under plurality and the other decides which party acquires seats under PR rules. Voters in such hybrid systems are offered an opportunity to split their votes for different political affiliations. Ticket-splitting in countries with mixed-member systems is thus different from that which occurs in the American context. Both Japan and Taiwan abandoned the time-honored single nontrans- ferable vote electoral rules in favor of the MMM systems in 1994 and 2005, respectively. The new institutional setting of both countries provides citi- zens with two ballots.' Voters in both countries may thus behave strategi- cally as the new institutions offer voters an opportunity to split their votes for candidates of different political affiliations. Parties of similar political positions may also plan their campaigns collectively in order to maximize their overall seat shares. This study examines how voters are expected to behave under the new systems, and, in particular, how they respond to political parties' strategic moves of maximizing electoral votes. To examine voters' electoral behavior under the new electoral rules, survey data collected in Japan and in Taiwan are employed, which include 196 Mixed-Member Electoral Systems in Constitutional Context data of the 2008 Taiwan's Election and Democratization Study (hereaf- ter TEDS2008L),2 fifteen waves of preelection rolling surveys conducted prior to Taiwan's 2012 combined elections,3 and Japan Election Studies (JES) III.4 This study proceeds in three parts. The next section reviews the literature on split-ticket voting, which offers a theoretical framework for analysis. In the ensuing sections, descriptive and multivariate analyses are conducted, with conclusions in the final section. The findings support the conventional wisdom that party identification is the primary determinant of electoral behavior. The institutional design of mixed-member systems, however, mitigates the effects of partisan loyalty as identifiers of minor political parties are more likely to be ticket-splitters, but they tend to sup- port candidates of the same political coalitions/alliances. Voters' ticket- splitting behavior may be further induced by coordinated nomination strategies of political parties with similar political positions. Split-Ticket Voting in MMM System More than half a century ago, Angus Campbell and his associates (1964) stated that: Few factors are of greater importance for our national elections than the lasting attachment of tens of millions of Americans to one of the parties. These loyalties establish a basic division of electoral strength within which the competition of particular campaigns takes place.... Most Americans have this sense of attachment with one party or the other. And for the individual who does, the strength and direction of party identification are facts of central importance in accounting for attitude and behavior. (67-68) Generally characterized as the "Michigan model," its characterization of party identification as the primary determinant of political behavior has subsequently dominated the academic research in the discipline (e.g., Green, Palmquist, and Schickler 2002; Miller and Shanks 1996). Although a number of studies have noted an increase in "independent voters" along with split-ticket voting over the subsequent decades (Burnham 1970, 123; Keith et al. 1992; Pomper 1967), more recent analysis has shown that the argument for the "decline of parties" in American politics was exagger- ated. Bartels (2000), for instance, demonstrates that the impact of partisan loyalties on presidential and congressional voting was strong in the elec- Split-Ticket Voting under MMM 197 tions held between 1952 and 1996 in the United States. Other recent stud- ies (Hetherington 2011; Stanley and Niemi 2010) have also shown that straight-ticket voting by Americans has increased. Although the effect of party identification on electoral choice varies by candidates' social char- acteristics (Campbell, Green, and Layman 2011), over time (Bartels 2000; Lewis-Beck et al. 2008), and across types of elections (Bartels 2000), party identification continues to play an important role in American electoral politics. In elections between candidates from competing parties, individu- als affirm their partisan loyalties by casting ballots for the candidates who share their party labels. While this conclusion is primarily based on find- ings in an American setting, empirical analysis shows that it is also appli- cable to other democratic countries such as Taiwan (Chu 1996; Fu 1996; Ho 1994; Liu 1994; Sheng 2002; Sheng and Chen 2003; Shyu 1991, 1992; Tsai 2008; N. Wu 1999; Yu 2004) and Japan (Akuto 1971; Kohei 1972; Kuo 2011; Richardson 1988; Rochon 1981). Although party labels serve as an important cue for individuals' voting decisions, the institutional framework within which an election operates may determine how much partisanship matters. In this context, the semi- nal work by Duverger (1959) becomes relevant. He argues that the SMD and the PR systems, in their pure forms, each provide different incentives to party elites and to voters. In a pure SMD election, the simple mechan- ics of the plurality rule create an effect of "underrepresentation" of the weaker parties because a larger party can win 100% of the seats available by winning half or even less than half of the votes cast in each district. As a result of this mechanical effect of the plurality rule, the percentage of seats of minor parties tends to be disproportionate to their percentage of the poll. Such a mechanical effect further introduces a psychological effect in voters' minds. As they realize that backing trailing candidates has a higher chance of becoming "wasted votes," supporters of these candidates in the SMD context are likely to transfer their votes to more viable candidates of major parties. In a similar vein, party elites behave strategically in the SMD system by not wasting resources on candidates who have little chance of winning. Recognizing the mechanical and psychological effects of the SMD system, minor parties or parties of similar political positions may also strategically choose not to nominate candidates in some or all of the SMDs lest they divide the pool of support and thereby lose the election. Such strategic decisions of nomination mean that the voters' choice menu in any given SMD tier is typically more limited than the one in the PR tier, further aggravating the effects of the SMD system. This logic leads to Duverger's 198 Mixed-Member Electoral Systems in Constitutional Context classic argument that "the simple-majority single ballot system favors the two-party system" and thus the number of parties will settle at two in a pure SMD electoral system (1959, 217). In a pure PR electoral system, however, there is little incentive for parties of similar political tendencies to fuse because the division does little or no harm to them. As the PR system puts a "powerful brake" to any tendency toward a two-party system, Duverger concludes that "proportional representation encourages a system of parties that are multiple, rigid, independent and stable" (1959, 204-5). The SMD and the PR systems in their pure forms thus offer differ- ent incentives to voters. Because an MMM system is a hybrid of both PR and SMD, voters in this system may behave strategically, leading to split- ticket voting. Various studies show ticket-splitting in such hybrid systems as evidence of voters' strategic reaction to electoral rules because they cast a greater number of votes for large parties in SMD races than in the PR tier (Bawn 1999; Cox 1997; Huang and Hsiao 2009; Reed 1999; Roberts 1988).5 As a result, identifiers with major parties tend to cast sincere votes in both SMD and PR segments by voting for their most preferred can- didates and parties. Supporters of minor parties may act strategically in district races by voting for the more viable candidates of major parties but cast a sincere PR vote for their preferred minor parties. While partisan identification influences individuals' vote choice, the electoral context may mitigate its effects. The tendency of ticket-splitting on the part of voters may be further induced by the strategic moves of party elites. In addition to fielding can- didates in SMDs in order to boost party votes in the PR tier by generat- ing the so-called contamination effect (Cox and Schoppa 2002; Ferrara and Herron 2005; Ferrara, Herron, and Nishikawa 2005; Herron and Nishikawa 2001; Huang 2007; Mizusaki and Mori 1998; Reed 2003),6 an approach frequently adopted by political parties in Japan and in Taiwan has been to coordinate candidate nomination in district races with other parties. Because there are no linkages between the SMD and the PR tiers in MMM systems, victory in district races directly changes the number of seats won by parties. Such electoral rules thus provide incentives to par- ties of similar political positions to coordinate their candidate nomination in SMDs lest they divide the pool of support (Ferrara and Herron 2005). Coordinated district races can also boost parties' PR votes as voters receive a clear voting cue from the electoral coalitions. Thus, parties of electoral coalitions can collectively maximize their seat shares through preelectoral pacts. Indeed, political parties in both Japan and Taiwan have coordinated their campaigns since electoral reforms were passed. In no district did Split-Ticket Voting under MMM 199 the Liberal Democratic Party (LDP) and the Clean Government Party (Komeito, CGP) of Japan both nominate candidates in SMD races in 2003 and 2005. Similarly, Taiwan's two major political parties, the Nationalist Party (Kuomintang, KMT) and the Democratic Progressive Party (DPP), also collaborated with their junior partners in the 2008 and 2012 legislative elections.7 The above discussion thus yields the following theoretical expectations to be examined in the subsequent sections. Specifically, it is hypothesized that partisan identification remains the primary determinant of vote choice in mixed-member systems but the framework of the electoral institutions mitigates its effect. Identifiers of major political parties are more likely to cast straight tickets while supporters for minor parties tend to split their ballots by casting a sincere PR vote and confine their defection to SMD races. As political parties may coordinate their campaign strategies, voters' ticket-splitting behavior may be further induced by such preelectoral pacts. Split-Ticket Voting in Japan and Taiwan Since MMM systems were adopted in the two countries, the 2003 and 2005 legislative elections were the third and fourth such contests held in Japan, whereas the 2008 and 2012 elections were the first and second held in Taiwan. The vote and seat shares by political parties in the elections are presented in tables 7.1a and 7.1b. Data in table 7.la show that the LDP in Japan, the ruling party since 1955, was able to continue its domination in the legislative elections. In 2003, the party garnered about 44% of the electoral support in SMD races and 35% of the PR votes, which were translated into a total of 237 legisla- tive seats. The opposition Democratic Party of Japan (DPJ) also acquired a substantial numbers of votes in both tiers and became the second largest political party in the Diet with 177 seats. The LDP advanced its politi- cal domination in the 2005 elections with a landslide victory by winning 296 of the 480 total legislative seats. Along with its political partner, the LDP-CGP governing coalition commanded a two-thirds majority in the lower house. The DPJ suffered a devastating loss in 2005, acquiring only 113 seats but continuing to be the largest opposition party. Overall, the party system in Japan in the mid-2000s can be characterized as heading toward a two-party system dominated by the LDP and the DPJ, or more precisely, a "governing coalition vs. opposition coalition" system. Indeed, observers have noted that the effective number of parties in Japan has 200 Mixed-Member Electoral Systems in Constitutional Context shown a downward trend during the postreform era but has never reached two. Smaller parties have been able to keep a presence in the Diet even though they cannot thrive (Jon 2009; Maeda 2008) and major political par- ties have maintained collaborative relationships with them. For instance, the CGP continued to be the partner of the LDP-dominated governing coalition even after the LDP's landslide victory in 2005. The DPJ has also maintained a loosely defined political alliance with the Social Democratic Party (SDP) through collaborative campaigns in some, but not all, SMDs. The only exception to this pattern is the Japanese Communist Party (JCP), which has continued its adherence to an unwavering political ideology and "going it alone" (Ferrara and Herron 2005) by fielding a large number of candidates in SMDs regardless of the chances of winning.8 TABLE 7.1A. Vote and Seat Shares by Political Party in Japan 2003 2005 Vote % Number of Seats (%) Vote % Number of Seats (%) SMD PR SMD PR SMD PR SMD PR LDP 43.9 35.0 168 (56.0) 69 (38.3) 47.8 38.2 219 (73.0) 77 (42.8) DPJ 36.7 37.4 105 (35.0) 72 (40.0) 36.4 31.0 52 (17.3) 61 (33.9) JCP 8.1 7.8 0(0.0) 9(5.0) 7.3 7.3 0(0.0) 9(5.0) CGP 1.5 14.8 9(3.0) 25 (13.9) 1.4 13.3 8(2.7) 23 (12.8) SDP 2.9 5.1 1(0.3) 5(2.8) 1.5 5.5 1(0.3) 6 (3.3) NCP 1.3 - 4(1.3) - - - - - PNP - _ - - 0.6 1.7 2 (0.7) 2 (1.1) NPN - _ - - 0.2 2.4 0 (0.0) 1(0.6) NPD - _ - - 0.1 0.6 0 (0.0) 1(0.6) Others 5.7 0.0 13 (4.3) 0 (0.0) 4.8 0.00 18 (6.00) 0 (0.0) Data source: Revised from Kuo (2011, 20-21). TABLE 7.1 B. Vote and Seat Shares by Political Party in Taiwan 2008 2012 Vote % Number of Seats (%) Vote % Number of Seats (%) SMD PR SMD PR SMD PR SMD PR KMT 53.5 51.2 57 (78.1) 20 (58.8) 48.1 44.6 44 (60.3) 16 (47.1) DPP 38.7 36.9 13 (17.8) 14 (41.2) 44.5 34.6 27 (37.0) 13 (38.2) NP - 4.0 - 0((0.0) 0.1 1.5 0(0.0) 0(0.0) PFP 0.1 - 0(0.0) - 1.1 5.5 0(0.0) 2 (5.9) TSU 1.0 3.5 0 (0.0) 0 (0.0) - 9.0 - 3 (8.8) NPSU 2.3 0.7 2 (2.7) 0 (0.0) 1.1 - 1(1.4) - Others 4.6 4.4 1(1.4) 0 (0.0) 4.9 4.9 1(1.4) 0 (0.0) Data source: Central Election Commission, Taiwan. Note: The six Aboriginal seats are not included in this table. Split-Ticket Voting under MMM 201 Taiwan's political landscape has been roughly divided into two politi- cal camps: the Pan-Blue alliance and the Pan-Green alliance, which are divided on the issues of national identity and cross-strait policy (Hsu and Lin 2009; Niou 2004; Wang and Liu 2004; Y. Wu 2004). The Pan-Blue alliance consists of the now ruling KMT, the People First Party (PFP), and the New Party (NP), while the Pan-Green alliance includes the oppo- sition DPP and the Taiwan Solidarity Union (TSU). In the first legisla- tive election in 2008 since the MA/M system was adopted, parties of the Pan-Blue alliance entered an electoral pact to coordinate their campaign. Also included in the agreements was the Non-Partisan Solidarity Union (NPSU). The two political parties of the Pan-Green alliance, however, failed to reach an agreement. They fielded their candidates separately in both the SMD and PR tiers and competed with each other for the support of Pan-Green identifiers. As the data in table 7.lb show, the KMT and the DPP, the two largest political parties in Taiwan, garnered almost all of the legislative seats in 2008 and none of the minor parties passed the required 5% threshold to receive apportioned PR seats. In the 2012 combined presidential and legislative elections, the KMT again collaborated with the NP and the NPSU in the legislative campaign. With a coordinated nomination strategy, the NP presented a party list in the PR tier and nominated only one SMD candidate while the NPSU nominated two SMD candidates without competing in the PR segment. The KMT fielded candidates in the remaining 71 SMDs and presented a complete party list. The PFP did not enter the 2012 Pan-Blue pact and ran its legislative campaign separately mainly due to the presidential bid by its chairman, James Soong, who ran against the incumbent president, Ma Ying-jeou of the KMT. On the part of the Pan-Green alliance, the DPP fielded 69 candidates for district races while the TSU only presented a party list in the PR tier without running any candidates in SMDs. Although the two parties did not enter a formal agreement, it appeared that the TSU and the DPP reached a tacit understanding on their campaign strategies as leaders of the TSU repeatedly called on Pan-Green identifiers to sup- port DPP candidates in SMD races but back the TSU with their party votes. The DPP chairwoman, Tsai Ing-wen, also indicated publicly that she would like to see the TSU garner more than 5% of the party votes so that the party could receive apportioned PR seats (Lee 2012). Similar to the 2008 electoral outcome, the KMT and the DPP had a lion's share of the legislative seats with the former as the biggest winner. While the PFP and the TSU were able to pass the 5% threshold and acquired two and three seats in the PR tiers, respectively, the electoral results of 2008 and 2012 202 Mixed-Member Electoral Systems in Constitutional Context demonstrate that minor political parties in Taiwan have difficulty winning SMD races without collaborating with major parties. Interestingly, as both Japan and Taiwan are in the process of consoli- dating their two-party systems, the vote shares of major and minor par- ties in SMD races are generally different from their PR vote shares. The divergent patterns of vote distributions at the aggregate level seem to show that some electorates in both countries split their tickets in the elections. Survey data from the JES III and the TEDS appear to confirm this finding. By cross-tabulating SMD and PR votes at the individual level, table 7.2a shows that, in the 2003 elections, 37.8% of the Japanese electorate cast a straight-ticket vote for the LDP in both PR and SMD tiers, while 26.6% of them voted sincerely for the DPJ in both segments. Even after other minor parties are included in the calculation, less than 70% of Japanese citizens were straight-ticket voters. Similar patterns can be found in the 2005 electoral results as a total of 68.5% of JES respondents supported candidates of the same political party in both the PR and SMD tiers. This means that roughly 30% of Japanese citizens were ticket-splitters in these two legislative elections. By comparison, the proportions of straight-ticket voters were higher in Taiwan's 2008 legislative elections as 77.7% of the TEDS respondents indicated that they supported the same political par- ties in both the PR and SMD tiers. Specifically, 55.2% and 22.5% of the electorate, respectively, were KMT and DPP straight-ticket voters in 2008. The 2012 elections, however, witnessed a 10% increase in ticket-splitters and less than 70% of the country's citizens cast straight tickets. While some of these ticket-splitters may be "nonopportunity voters"9 due to a lack of SMD nominations of their preferred political parties (Benoit, Giannetti, and Laver 2006; Ferrara, Herron, and Nishikawa 2005, 87), roughly 20% to 30% of Japanese and Taiwanese voters split their tickets between their preferred parties in the PR list and candidates of other political affiliations in SMD races. TABLE 7.2A. Straight Ticket Voters in the Electorates (in percentages) Japan Taiwan 2003 2005 2008 2012 LDP 37.8 38.0 KMT 55.2 50.6 DPJ 26.6 26.4 DPP 22.5 17.0 CGP 0.8 0.7 TSU 0.0 - SDP 0.8 0.2 PFP - 0.4 JCP 3.2 3.2 NPSU 0.0 - All Straight Tickets 69.2 68.5 All Straight Tickets 77.7 68.0 Data source: JESIII, TEDS2008L, TEDS2012-T, Huang (2011), and Yu (2011). Split-Ticket Voting under MMM 203 Comparing with ticket-splitting under the MMP systems in Germany and New Zealand, the patterns of split-ticket voting in Japan and in Tai- wan are not unique. Indeed, the proportion of German ticket-splitters has increased from 4.3% in 1953, when the country adopted the MMP system, to 22.1% in 1998 (Gschwend, Johnston, and Pattie 2003; Kinge- mann and Wessels 2001). New Zealand also witnessed 37% of its voters splitting their ballots in the country's first parliamentary election under the MMP rules (Johnston and Pattie 2002; Barker et al. 2001). The extent of ticket-splitting in Japan and in Taiwan is thus comparable with that in other countries that employ mixed-member electoral systems. It is important to note that ticket-splitters in both countries tend to vote for ideologically proximate candidates. As table 7.2b shows, more than 75% of CGP and 44% to 60% of SDP supporters in the PR tier, respec- tively, voted for Japan's LDP and DPJ candidates in SMD races. About 90% of the NP's and 60% to 80% of the TSU's PR supporters in Taiwan TABLE 7.2B. Vote Choices between PR and SMD Ballots 2003/2005 SMD Ballot Japan LDP DPJ CGP SDP JCP n LDP 82.7 6.5 2.4 0.5 0.8 660 2003 DPJ 15.2 72.2 0.6 4.1 2.6 532 PR Ballot CGP 76.6 10.2 8.8 0 0.7 137 SDP 18.8 43.8 0 25.0 8.3 48 JCP 8.8 20.6 0 2.9 67.7 68 LDP 83.5 8.2 3.6 0.2 0.9 563 2005 DPJ 11.6 82.3 0.3 0 1.3 396 PR Ballot CGP 78.8 6.6 6.6 0 1.5 137 SDP 16.7 60.4 0 6.3 12.5 48 JCP 9.7 16.1 0 1.6 64.5 62 2008/2012 SMD Ballot Taiwan KMT DPP TSU PFP NPSU n KMT 89.8 4.3 0 - 2.7 625 2008 DPP 7.2 86.7 0.8 - 0.4 264 PR Ballot NP 90.0 1.7 0 - 1.7 60 TSU 21.9 62.5 0 - 3.1 32 NPSU 33.3 16.7 16.7 - 0 6 KMT 89.9 6.1 - 0.6 1.4 1,953 DPP 14.3 81.6 - 0.4 0.6 723 2012 NP 89.7 5.1 - 1.3 0 78 PR Ballot PFP 62.3 25.2 - 4.3 1.3 320 TSU 14.2 80.9 - 0.3 0 325 Data source: JESIII, TEDS2008L, TEDS2012-T, Huang (2011), and Yu (2011). Note: Cells are row percentages; figures are listed for selected political parties. 204 Mixed-Member Electoral Systems in Constitutional Context cast their SMD ballots for KMT and DPP candidates, respectively. These findings indicate that ticket-splitting in Japan and Taiwan tends to occur within coalitions rather than across coalitions. Indeed, party identification represents a psychological attachment to a political group and manifests individuals' political ideologies and positions on a host of issues. To be con- sistent with their partisan loyalty, ticket-splitters thus tend to back those within the partisan coalition. That said, one cannot ignore the fact that there are a significant number of voters in Japan and Taiwan that split their tickets across coalitions. Table 7.2b shows that about 6% to 8% of LDP and 12% to 15% of DPJ support- ers in the PR tier, respectively, voted for DPJ and LDP candidates in SMD races. Similarly, 4% to 6% of the KMT's supporters and 7% to 14% of the DPP's supporters cast their SMD ballots for candidates of the other major party. These findings show that strategic thinking is not the only factor involved in ticket-splitting. Voters may support a candidate due to personal ties or the candidate's service record to the constituency regardless of his/ her partisan affiliation. The tendency to cast a personal vote is particularly strong in Japan and in Taiwan because both countries previously employed the single nontransferable vote system for legislative elections, which is notoriously characterized by personal ties between candidates and voters. The parallel institutional features of the MMM system discussed in chap- ter 1 also provide incentives for parties to encourage their district candi- dates to launch more personalistic campaigns (Moser and Scheiner 2004). Despite these historical and institutional factors contributing to personal votes, the overall voting patterns in both Japan and Taiwan lend support for the proposition that partisan identification is the primary determinant of vote choice but electoral institutions mitigate its effect. Our theoretical expectations are that identifiers with major parties tend to cast sincere votes in both SMD and PR segments for their most pre- ferred candidates and parties. Supporters of minor parties may act strate- gically by casting a sincere vote in the PR tier but then transferring their support to more viable SMD candidates. To test these hypotheses, tables 7.3a and 7.3b cross-tabulate individual respondents' PR and SMD votes after their partisan loyalty is identified. The empirical results support the above theoretical expectations. Specifically, 56% to 60% of LDP support- ers and about 66% of DPJ identifiers in both the 2003 and 2005 elections cast straight tickets for their preferred parties and the affiliated candidates, whereas 59% to 67% of CGP supporters and 23% to 30% of SDP identi- fiers were ticket-splitters. While supporters of major parties in Japan had a greater tendency to cast straight tickets, identifiers with minor parties Split-Ticket Voting under MMM 205 are more likely to vote sincerely in the PR tier for their most preferred parties but back more viable SMD candidates of the major party of the same political coalition. Apparently, the psychological effects discussed by Duverger were in operation in the mind of Japanese voters. The analysis of TEDS data yields similar findings in Taiwan. Because both the NP and PFP-two minor parties of the Pan-Blue alliance entered a preelectoral pact with the KMT in the 2008 elections, and thus they either did not nominate any candidates in district races or had their candidates run under the KMT banner. While almost all of the NP identi- fiers continued to be nonopportunity voters in the 2012 election and split their tickets in the PR and SMD tiers, some of the PFP identifiers had the opportunity to be straight-ticket voters as the PFP nominated 10 candi- dates in district races. Nevertheless, 35% and 15% of PFP identifiers in TABLE 7.3A. Vote Choices between PR and SMD Ballots by Party Identification in Japan Voter's party ID PR vote LDP LDP DPJ CGP SDP JCP CGP LDP DPJ CGP SDP JCP DPJ LDP DPJ CGP SDP JCP SDP LDP DPJ CGP SDP JCP JCP LDP DPJ CGP SDP JCP LDP 56.0 5.2 5.5 0.4 0.4 3.5 1.1 59.3 3.9 7.7 0.3 0.3 3.9 3.9 1.9 7.7 3.5 1.8 2003 SMD vote DPJ CGP SDP 3.8 1.7 0.2 11.7 0.1 0.5 0.2 0.4 0.2 JCP LDP 0.4 60.2 0.4 4.5 0.1 6.8 0.6 0.5 0.6 3.9 2005 SMD vote DPJ CGP SDP 4.3 2.2 0.2 8.5 0.5 0.3 0.3 0.2 1.3 JCP 0.5 0.2 0.2 2.3 9.3 1.2 1.8 65.8 0.3 0.9 1.2 19.2 23.1 1.9 15.8 1.2 8.1 0.3 0.3 3.9 5.1 66.7 6.4 7.5 1.2 4.4 0.9 0.3 0.6 1.2 2.8 66.9 0.3 4.1 1.3 7.7 0.9 0.3 3.0 0.3 0.9 0.3 0.3 0.3 0.3 0.3 0.9 1.9 6.1 1.9 1.9 27.3 3.0 11.5 1.9 3.9 1.8 3.5 1.8 56.1 3.0 30.3 7.4 11.1 6.1 6.1 6.1 1.9 1.9 1.8 1.9 3.7 5.6 48.2 1.8 5.3 Data source: JESIII. Note: Figures given in the cells are total percentages. Percentages are presented only for selected political par- ties, thus the sum of percentages does not equal 100%. 206 Mixed-Member Electoral Systems in Constitutional Context 2012 decided to cross the party line and supported KMT and DPP can- didates in SMD races, respectively. As in Japan, the psychological effect appeared to be in place at least in some of the PFP identifiers. Probably the most instructive finding is the voting patterns of TSU identifiers. Recall that Pan-Green parties failed to reach a preelectoral pact ahead of the 2008 elections. As a result, the TSU fielded a number of SMD candidates with a complete PR list to compete with the DPP for the support of Pan-Green identifiers. The electoral outcome in 2008 was disastrous for the party as it did not win any district race and failed to pass the 5% threshold in the PR tier. Recognizing the mistake they made, TSU leaders decided to collaborate with the DPP in the 2012 elections. This TABLE 7.31B. Vote Choices between PR and SMD Ballots by Party Identification in Taiwan Voter's 2008 SMD vote 2012 SMD vote party ID PR vote KMT DPP TSU KMT DPP PFP KMT KMT 82.2 3.6 78.6 3.9 0.5 DPP 0.4 0.8 1.0 2.1 NP 4.1 0.2 2.5 0.1 0.1 PEP - - 4.5 1.0 0.1 TSU 0.2 0.4 0.6 NP KMT 28.6 52.1 2.1 DPP NP 50.0 25.0 4.2 PEP - - 2.1 2.1 TSU 3.6 2.1 PFP KMT 50.0 6.3 14.3 1.4 DPP 0.7 3.4 NP 18.8 2.0 PEP - - 34.7 15.0 6.1 TSU 6.3 2.0 6.1 DPP KMT 3.1 0.9 5.3 1.9 0.1 DPP 3.1 82.5 0.5 5.8 52.7 0.2 NP 0.5 0.2 PEP - - 1.7 2.3 TSU 2.2 2.7 20.1 0.1 TSU KMT 17.4 4.4 DPP 21.7 6.7 28.9 NP PFP 4.4 TSU 13.0 17.4 8.9 46.7 Data source: TEDS2008L, TEDS2012-T, Huang (2011), and Yu (2011). Note: Figures given in the cells are total percentages. Percentages are presented only for selected politi- cal parties, thus the sum of percentages does not equal 100%. Split-Ticket Voting under MMM 207 strategy appeared to be working for the TSU. By not competing with an ideologically proximate major party in district races, the TSU successfully consolidated the party votes from its own followers. Interestingly, the TSU leaders' call on Pan-Green identifiers to support DPP candidates in SMD races but back the TSU in the PR also drew 20% of the DPP identifiers who transferred their party votes to the TSU. Party elites' explicit mes- sage of encouraging such ticket-split voting may be responsible for DPP identifiers' support for the TSU in the PR tier. Taiwan's single national 5% threshold also makes this type of strategic voting likely by comparison as voters can more easily figure out if a small party is close to the threshold.'0 As a result, the TSU garnered about 9% of the party vote and received three seats in the PR tier. As shown in table 7.3b, 29% of TSU identifiers cast a straight party vote for the DPP in the 2012 election. Unlike other minor party iden- tifiers who split their ballots by casting sincere PR votes for their most preferred parties but voting for viable candidates in the district races, the psychological effect appeared to be in full swing here. Close to one-third of TSU identifiers worried that their votes would be wasted in the SMD as well as in the PR tiers if they supported a minor party. These voters' strategic calculation was probably due to the 5% legal threshold of Tai- wan's PR segment. The TSU's disastrous outcome in 2008 might have also reinforced the Duvergerian gravity in both SMD and PR segments of the 2012 elections. Similarly, the same logic seems applicable to NP identifiers because more than half of them cast a straight party vote for the KMT in the 2012 election while only 25% of NP identifiers split their ballots and cast a sincere vote in the PR tier. The voting patterns of TSU and NP iden- tifiers lend additional support for the hypothesis that the effect of partisan identification on electoral choice may be moderated by the institutional framework of MAM systems. More important, such a mitigation effect is more apparent to identifiers of minor parties than to those of major parties. That said, it is important to note that some citizens in Japan's and in Taiwan's legislative elections were nonopportunity voters. As table 7.4 shows, almost all of the political parties in the two countries' elections pre- sented a party list to compete in the PR tier. None of them, however, made nominations in all district races, except the JCP in Japan's 2003 election. The lack of district candidates was particularly acute for such minor parties as the CGP and the SDP in Japan and the NP, the PFP, the NPSU, and the TSU in Taiwan. While all voters in mixed-member electoral systems have the option of ticket-splitting, voters in these districts were forced to do so because their most preferred party failed to provide a choice on both PR 208 Mixed-Member Electoral Systems in Constitutional Context and SMD ballots. To avoid overestimating ticket-splitting, a proper analy- sis of electoral behavior in Japan and in Taiwan thus requires attention to the voting patterns of these nonopportunity voters. In general, the above analysis shows that 20% to 30% of voters split their tickets in the respective elections in Japan and in Taiwan. The voting patterns in both countries seem to be consistent with the theoretical expec- tations that partisan identification continues to be the primary determinant of vote choice in MMM systems but the electoral institutions mitigate its effect. Thus, identifiers of major political parties tend to cast straight tick- ets while supporters for minor parties are likely to be sincere in their PR votes but vote for viable candidates in SMD races. When they cross the party line and vote for candidates of different political affiliations, split- ticket voting in both countries tends to occur within coalitions rather than across coalitions. As political parties may coordinate their campaign strate- gies, voters' ticket-splitting may be further induced by such preelectoral pacts. Because some of these ticket-splitters were nonopportunity voters due to a lack of SMD nominations of preferred political parties, a multi- variate analysis of JES III and TEDS survey data is in order to ensure that the above findings are not spurious. Data and Methods To conduct the multivariate analyses, the dependent variable, Vote Choice, is created based on voters' responses if they cast a straight-ticket or split their ballots in the elections. Due to the different political landscapes of Japan and Taiwan, the variable is coded with different labels but each has five TABLE 7.4. "Nonopportunity" Voters: Number of Districts with a Party Nominee Japan Taiwan 2003 2005 2008 2012 SMD PR SMD PR SMD PR SMD PR LDP 277 11 290 11 KMT 69 Yes 69 Yes CGP 10 11 9 11 NP 0 Yes 1 Yes DPJ 267 11 289 11 PFP 1 No 10 Yes SDP 61 11 37 11 NPSU 5 Yes 2 No JCP 300 11 273 11 DPP 69 Yes 69 Yes TSU 13 Yes 0 Yes Data source: Central Election Commission, Taiwan; Elections and Citizen Participation website: http:// www.asianelections.org/cubekmcht/ezcatfiles/cust/img/img/144/196798271.pdf; http://www.asianelec tions.org/cubekmcht/ezcatfiles/cust/img/img/144/467755124.pdf. Note: Japan has 300 SMDs and 11 PR districts. Taiwan has 73 SMDs and one PR district. Split-Ticket Voting under MMM 209 categories. Because the Japanese political system can be roughly character- ized as "governing coalition vs. opposition coalition," citizens' vote choices in Japan include (1) governing coalition straight-ticket, (2) opposition coalition straight-ticket, (3) governing coalition split-ticket, (4) opposi- tion coalition split-ticket, and (5) other forms of split-ticket voting." To be consistent with Taiwan's "Pan-Blue vs. Pan-Green" political landscape, the categories of the island citizens' vote choice consist of (1) Pan-Blue alliance straight-ticket, (2) Pan-Green alliance straight-ticket, (3) Pan-Blue alliance split-ticket, (4) Pan-Green alliance split-ticket, and (5) other forms of split- ticket voting. Table 7.5 presents the frequency distributions of each voting type in Japan and in Taiwan, respectively. As indicated earlier, the voting behavior of Japanese citizens appears to be more stable because there are few variations in the distributions of various vote choices. Taiwanese citi- zens' voting patterns appear to be more volatile as there is a 10% shift from straight-ticket to split-ticket voting. Such changes are particularly visible in the support for the Pan-Green alliance, which, as indicated earlier, was attributed to the coordinated electoral campaigns of the DPP and the TSU. While data from Japan and Taiwan are analyzed separately, several common independent variables are created so that cross-country compari- sons can be made. Because partisan attachment is considered the primary determinant of an individual's vote choice, a series of dummy variables of respondents' party identifications are created separately for each country. To assess Japanese citizens' partisan loyalties, the following variables are coded with 1s for respondents in the relevant categories and otherwise with Os for nonpartisan voters as the baseline group: Liberal Democratic Party (LDP), Democratic Party of Japan (DPJ), Clean Government Party (CGP), Social Democratic Party (SDP), and Japanese Communist Party (CP). Corre- sponding variables of Taiwanese citizens' party identification are National- TABLE 7.5. Distribution of Citizens' Vote Choices Typologies Japan Taiwan 2003 2005 2008 2012 Governing straight 38.6 38.8 Pan-Blue straight 55.3 51.0 Opposition straight 30.6 29.9 Pan-Green straight 22.6 17.0 Governing split 9.3 10.4 Pan-Blue split 7.3 8.7 Opposition split 5.3 5.3 Pan-Green split 2.2 7.6 Other split 16.2 15.8 Other split 12.6 15.7 Total (n) 1,445 1,236 Total (n) 1,014 3,466 Data source: JESIII, TEDS2008L, TEDS2012-T, Huang (2011), and Yu (2011). 210 Mixed-Member Electoral Systems in Constitutional Context ist Party (KMT), Democratic Progressive Party (DPP), New Party (NP), People First Party (PFP), and Taiwan Solidarity Union (TSU). As indicated earlier, not all political parties in Japan's and in Taiwan's past legislative elections nominated candidates competing for SMD seats. Voters in such districts where their most preferred parties did not nomi- nate candidates would not even have the opportunity to cast straight tick- ets. In Taiwan's 2008 legislative elections, for instance, the DPP did not nominate candidates in four districts while the TSU did not do so in 60. In Taiwan's 2012 legislative elections, the TSU did not field any candidates in district races. Similarly, in Japan's 2003 and 2005 elections the LDP coordinated its nominations with the CGP and did not make nominations in those districts where the CGP candidates ran. The DPJ and the SDP also coordinated their nominations in some SMDs lest they compete with each other. The only political party that did not enter into electoral pacts with other parties was the JCP as its candidates ran in all 300 SMDs in the 2003 elections and in 273 SMDs in 2005. To avoid overestimating ticket- splitting, three dummy variables are incorporated to control the effect of "nonopportunity" supporters in Japan: nomination for DPJ, nomination for LDP, and nomination for SDP Similarly, three corresponding dummy vari- ables are included to control the effect of "nonopportunity" supporters in Taiwan: nomination for KMT, nomination for DPP, nomination for TSU, and nomination for PFP" These dummy variables are coded as is if the relevant political parties nominated candidates in the SMDs and Os if the parties did not field candidates. Several demographic factors are included as control variables. Previous studies find that education increases an individual's ability to understand abstract and intangible subjects like politics. Such abilities heighten indi- viduals' interest in politics, equip them with better knowledge of demo- cratic norms and procedures, give them greater skills in working with com- plex rules, and offer them the capacity of identifying local and national leaders (Huang, Yu, and Hsiao 2011; Karp et al. 2002; Nie, Junn, and Stehlik-Barry 1996; Wolfinger and Rosenstone 1980). Well-educated vot- ers are more likely to possess the political awareness and ability to under- stand the institutional design of MMM systems and the strategic payoffs of each menu choice in the PR and the SMD tiers. They are more likely to split their votes for candidates of different political affiliations due to their political sophistication. To tap respondents' education attainment, the fol- lowing dummy variables are included: university degree (and above), junior college degree, and senior high school degree. They are coded as is for those who are in the relevant category and Os otherwise, with respondents having Split-Ticket Voting under MMM 211 a junior high school degree and below as the reference group. To control the effects of gender and age, two variables are created: Male is a dichoto- mous variable, which is coded as 1 according to respondents' gender, and Age is measured by the number of years since birth. Because the dependent variable, Vote Choice, has five nonordered cate- gories, a multinomial logit model is used (Glasgow and Alvarez 2008; Long 1997; Long and Freese 2006). Specifically, the regression model takes the form of InSmlb(x-lfl -I x) m b, for m = 1 to]- 1, Pr(y=b x) where b is the base category for comparison; lnfmib(x) is the conditional odds of having other responses relative to the base category; x is a vector of independent variables; P is a vector of regression parameters; and]- is the number of categories in the dependent variable. Multivariate Analysis of Split-Ticket Voting in Japan and Taiwan Using "governing coalition straight-ticket" and "Pan-Blue alliance straight-ticket," respectively, as the baseline for comparison in analyzing Japanese and Taiwanese citizens' voting behavior, the variable of Vote Choice is regressed on the aforementioned independent variables in a multivariate context. The results in tables 7.6a to 7.6d show that most coefficients asso- ciated with partisan identification, especially those of major political parties (the LDP and DPJ in Japan and KMT and DPP in Taiwan), are statistically significant. These findings show that partisan loyalty is the most important factor in determining citizens' electoral choice in both Japan and Taiwan even after the effects of nonopportunity voters and citizens' demographic characteristics are controlled for. Most important, data in the tables con- firm the previous finding that identifiers of major parties are more likely to cast sincere straight-tickets whereas those of minor parties tend to engage in ticket-splitting voting within their respective coalitions/alliances. To illustrate, table 7.7 presents the predicted probabilities for Japanese and Taiwanese voters, respectively, with selected conditions. As table 7.7 shows, an average LDP or DPJ male identifier's predicted probabilities of casting a straight-ticket for candidates of his own coalition are between 0.7 and 0.8 when his most preferred party fielded SMD candidates in the 2003 and 2005 elections. The likelihood for this voter to split the ticket was rather small with a probability of less than 0.1. However, an identifier of minor 212 Mixed-Member Electoral Systems in Constitutional Context political parties like the CGP and SDP has a higher tendency of splitting the ticket within the coalition, with a probability between 0.3 and 0.7 in both elections with or without SMD nominations by his most preferred parties. Table 7.7 reveals a similar voting pattern in Taiwan, and it is instructive regarding the effects of preelectoral pacts among political parties. First, like Japanese voters, an average male KMT or DPP identifier in the 2008 elections had a rather high probability (about 0.85) of casting a straight- ticket but a miniscule likelihood of splitting his ballots when the most pre- ferred parties had candidates in district races. Identifiers of minor parties like the NP, which engaged in a preelectoral pact, or the PFP, which had its candidates run under the KMT banner, had higher probabilities of either splitting the ticket within the alliance or casting a Pan-Blue straight-ticket. Parties of the Pan-Green alliance did not coordinate their campaigns in 2008 and TSU identifiers apparently had a difficult time making a decision as their predicted probabilities of voting were evenly spread across various options. In the 2012 elections, the PFP dropped out of the Pan-Blue pact and PFP identifiers, like their CGP and SDP counterparts in Japan, had a higher probability of splitting the ticket with or without SMD nomina- tions by the party as the majority of them cast their SMD ballots for KMT candidates. Probably due to the coordinated campaign by the Pan-Green alliance in 2012, an average TSU identifier had a higher probability of splitting the ticket and voting for the TSU in the PR tier and for DPP candidates in district races. For the same reason, an average DPP supporter had a lower likelihood of casting a Pan-Green straight-ticket in 2012 than in 2008 as they split their support for DPP district candidates and for the TSU in the PR tier. Along with the findings in table 7.3b, the empirical evidence thus shows that the coordinated campaign strategy in the 2012 elections by the Pan-Green alliance consolidated TSU identifiers' support in the PR tier and also drew party votes from DPP supporters. Apparently, voters' ticket-splitting behavior can be further induced by the campaign strategies of political parties. Conclusions Characterized as hybrids that combine both PR and SMD rules, mixed- member systems have been adopted in many countries during the past two decades. Because the institutional arrangements of such electoral systems provide citizens with an opportunity to split their votes for candidates of TABLE 7.6A. Multinomial Logit Model of Straight- or Split-Ticket Voting in Japan's 2003 Elections Opposition coalition Governing coalition Opposition coalition straight-ticket split-ticket split-ticket Other split-ticket /3(S.F.) / (S..) /3(.) (S.F.) Part, identification (other parties & Ind. =0) LDP -1.28 ** (0.23) -0.12 (0.36) -2.45 **(0.45) -0.57 * (0.25) DPJ 3.14 (0.36) 0.74 (1.09) 1.98 ..(0.46) 2.04 . (0.39) CGP 1.35 (0.81) 3.45 ** (0.49) 0.36 (0.84) 0.55 (0.51) SDP 2.55 ** (0.78) 2.25 (0.98) 3.66 (0.83) 2.35 (0.82) JCP 2.91 ** (0.64) _ 201 ** (0.77) 0.99 (0.80) Nomination in SYID DPI (no nomination=0)) 1 .87 ** (0.40) 0.16 (0.39) -0.90 * (0.42) -0.78 ** (0.28) LDP (no nomination=O) -1.60 ** (0.38) -2.38 r (0.40) -2* * (0.50) -2.06 r* (0.37) SDP (no nomination=0) -0.22 (0.27) -0.04 (0.33) 0.18 (0.37) -1.27 ** (0.30) Education (Junior highi and belowv =0) Senior high school 0.27 (0.23) 0.03 (0.29) 0.62 (0.44) 0.35 (0.23) Junior college 0.60 (0.32) 0.25 (0.42) 1.28 " (0.55) 0.37 (0.33) LUniversitv and above 0.43 (0.31) 0.08 (0.43) 1.22 * (0.53) 0.31 (0.32) Male (Femalk=0) 0.43 * (0.18) C).18 (0.24) C).12 (0).30) 0.32 (C).18) Age -0.03 ** (0.01 ) -0.01 (0.01) -0.02 (0.01 ) -0.02 **(0.01) Constant 0.49 (0.74) 0.73 (0.89) 1.50 (1.04) 2.98 **(0.68) M'odel Information: ni = 1,432; Log Likelihood = 1,471.6477; LR x2 = 1,067.95; df = 52; p 3.0 until 1993. As for the MMP period, there was a major spike in Nv, by all three mea- sures in 1996, as we might expect. We might say there was substantial con- tamination from the party vote to the district vote, as the average district sees NV rise to 3.3. However, there is also an apparent reduction of linkage, 268 Mixed-Member Electoral Systems in Constitutional Context suggesting district patterns became more distinct than they had been pre- reform. We see evidence for this in the fact that the national aggregate N based on candidate votes was closer to that based on party-list votes than it was to the district mean. Then, with each passing election under MMP, there has been a down- ward trend in Nv based on candidate votes (both aggregate and district mean), even in years when N based on party votes turned upward, as it did in 2002 and, more modestly, in the 2005-11 period. Strikingly, the district mean Nv is lower in the most recent three elections under the MMP era than it was in any election in the later SMP period of 1978-93, except for 1987. This is, of course, evidence against one variant of the contamination thesis, because a case could be made that the average single-seat district is more "Duvergerian" under the proportional mixed-member system than it was during a time when such districts were the only game in town. The overall decline in fragmentation from its high point with the first MMP election suggests that the dominant trend is not so much contami- nation as it is the emergence of separate vote patterns in the two tiers.36 This is suggested most especially by the three recent elections when N on party votes has trended modestly upward again, but that based on elector- ate votes has been lower than in most of the elections of the latter prere- form period. It is worth noting that the pattern we see in figure 1 in which the two tiers showing a distinct "psychological" effect (in Duverger's terms) need not be typical of MMP. Compare figure 9.2, which replicates what we saw in figure 9.1, only for the case of Germany since 1953. Before unification in 1990, the mean district N was generally below 2.5, and hence quite "Duvergerian." Moreover, there is little difference between the aggregate NV for either nominal or list votes, suggesting little ticket-splitting-a find- ing consistent with literature that directly addressed the phenomenon (e.g., Fisher 1973). In the 1990s and especially in the 2005 and 2009 elections, on the other hand, the mean district has become notably more fragmented. All three indicators have turned sharply upward, and the somewhat larger gap between nominal and list votes implies more ticket-splitting as well (confirmed in a study of the 1998 election by Gschwend, Johnston, and Pattie 2003), but not-critically-to the extent of keeping anything like a Duvergerian equilibrium in the single-member districts. In fact, Ger- many's mean district N has not been below 2.5 since 1983, and rose to 2.9 in 2005 and 3.5 in 2009. Thus, at least over several recent elections, the two MMP systems seem to be diverging. Germany seems to be experiencing contamination from the PR tier to the nominal, whereas in New Zealand Political Consequences of New /ealands Mi MP System 269 (0 mfn C 0 0 CY) N 1950 1960 1970 1980 1990 2000 2010 Year - Aggregate candidate votes - -- - - District mean Aggregate party list votes Fig. 9.2. Effective number of parties over time in Germany the two tiers are showing distinct patterns-two-party dominance in the single-seat districts, but considerable third-party voting on the PR vote." Dual Candidacy and Personal-Vote Contamination Effects As we noted previously, some variants of the contamination thesis imply that large parties benefit from the "personal vote" cultivated by their incumbent MPs, and that this results in those parties that win many plurality races also performing especially well on the party-list vote. We have already cast some doubt on this form of contamination, by observing a recent growth in the gap between N\V based candidate and party votes." There are other possible forms of cross-tier contamination, however, which we take up in this section. In particular, the practice of dual candidacy may create impor- tant cross-tier contamination effects. We could observe contamination on the intraparty dimension, meaning how parties and candidates relate to one another and how much incentive legislators have to cultivate con- nections to a locality (Shugart 2005). Specifically, intraparty contamina- tion could mean either district-level electoral competition affording less local representation and accountability than a pure SMP system, or the list tier becoming "personalized" and "localized" in a way that would be 270 Mixed-Member Electoral Systems in Constitutional Context unusual if the system were entirely a list proportional system (assuming lists are closed, as they are in New Zealand). At least in the New Zealand context, the former direction of possible contamination-weakening of district-level incentives on politicians-is controversial and perceived as normatively "bad" due to the perceived illegitimacy of dual candidacy. The other direction of contamination likely would be considered "good" if it were perceived to be functioning. The consideration of how contamination works on the intraparty dimension in New Zealand is tied up in the issue of dual candidacy-the right of candidates to stand for office simultaneously in a district and on a party list. In New Zealand, as in Germany and Japan but unlike in Taiwan, dual candidacies are allowed. In Germany, there seems little controversy over the practice, in contrast to the deep controversy found in Japan, where the "best loser" provision by which parties may choose to rank their lists may exacerbate the perceived illegitimacy of dual candidacy. In New Zealand, the major parties, Labour and National, have tended to have a majority of their candidates run in both tiers simultaneously; in recent elections, the percentage has been 70% or higher. The Green Party likewise has dually nominated most of its candidates, including 97% of them in 2011. The mic- roparties' behavior has been somewhat variable. (Data from Vowles 2012.) Dual candidacy has increasingly come under scrutiny and criticism. In fact, it was explicitly one of the provisions subject to the review of MMP following the referendum of 2011, under the Electoral Referendum Act of 2010. A case can be made that dual candidacy can be an efficient and desir- able feature of a mixed-member system-"good contamination," in a sense. Allowing dual candidacy means that parties have an incentive to nominate high-quality personnel in districts that they may have little chance of win- ning. If dual candidacy were banned, a party might be disinclined to risk wasting a good candidate in a likely losing district contest, and instead ensure the candidate a place in parliament via a good rank on the party list. This risk would be especially acute for parties that expect to win seats only, or primarily, via the list. For instance, a party like the Greens, were there no dual candidacy, would be unlikely to nominate its best candidates in districts, because it depends on the party vote to gain seats. As the New Zealand Electoral Commission put it in its final report from the MMP review: "Without dual candidacy, MMP elections would be poorer con- tests" (Electoral Commission 2012, 9). A positive feature of dual candidacy is that it may encourage greater local representation than either a pure SMP or PR system would do. If MPs elected via the list set up local offices in the constituencies they con- Political Consequences of New Zealand's MMP System 27 1 tested but lost, then voters have representatives that they can contact for assistance from more than just the party that won the most votes in their area. Evidence from the parliament elected in 2011 appears to bear this out. Eleven dual-listed National Party MPs who won their seat via the party list maintain electorate offices and often identify themselves by titles such as "List MP based in [electorate name]." Labour has eight list MPs who follow a similar strategy of maintaining an electorate office.39 Even the Greens, despite not coming close to winning an electorate, maintain several "Electorate Offices" although in fact the offices are shared between multiple MPs and the offices are located in just a few cities where the party has its greatest strength. This pattern fits a party that is less geographi- cally focused in its platform, but that has mainly an urban constituency. As for the fourth largest party, New Zealand First, in 2011 all but the leader, Winston Peters, were dual nominees and all but two of their MPs maintain electorate offices.40 The biggest controversy over dual candidacy stems from incumbents losing their district race yet remaining in parliament via the party list. That these legislators are sometimes referred to by names like "zombies" or "back-door MPs" calls attention to their perceived illegitimacy. Despite the controversy, there is little objective evidence that it is a major phe- nomenon in the New Zealand experience with MMP. For example, Vowles (2012) shows that the overall turnover of MPs from one election to the next actually has been greater under MMP than it was under SMP. More- over, from 1999 through 2011, only once has the percentage of incumbent MPs defeated in a district race but remaining in office due to the list been above 10% of all district MPs (in 2005); it was only 2.9% in 2011. More- over, most of the district losers who remain by winning a list seat retire after just one more parliamentary term (Vowles 2012). The controversy over dual candidacy thus does not seem based in the actual results of elections in New Zealand. Moreover, if New Zealand- ers wanted to remove the possibility of district losers continuing, a simple "incumbent defeat assurance" provision could be inserted into the law as follows: a losing district incumbent whose own candidate vote trails his or her party's vote in the district is skipped on the list. In this way, if the MPs in question lost because their party became less popular (which will often be a nationwide swing, and not a specifically local effect) they can retain their seat if they are sufficiently valued by their party as to have a list rank that permits their election. However, if they lose because they are less popular than their party, they are done-the list can't "save" them from rejection by their electorate. Such a proposal is discussed in the Electoral Commission's 272 Mixed-Member Electoral Systems in Constitutional Context online Proposals Paper during the MMP review, although not specifically recommended.4' In the end, the Commission recommended no change to the dual-candidacy provision. Conclusion Fifteen years after the switch from SMP to MMP, not all New Zealanders are totally convinced of the merits of MMP as evidenced by the modest majority obtained by the current system at the referendum of 2011. The Royal Commission of 1985-86, which originally proposed the MMP sys- tem, certainly anticipated many of the well-known mechanical effects of MMP-particularly the high degree of proportionality-as the German example was available as a reference. Yet little did anybody anticipate that, beyond the mechanical effects, the switch to MMP would result in some features of New Zealand poli- tics that are distinctive. For example, the effective number of parties in the country has actually turned out to be somewhat lower than in Ger- many, despite the broad similarity of the systems. In recent years, the two cases have diverged, with Germany becoming more fragmented in both the nominal and list tiers. By contrast, in New Zealand it seems that the tiers may be diverging from one another, as the two traditional parties, Labour and National, continue to dominate the district competition while various other parties combine for 20-25 % of the party vote and a compa- rable share of the seats. New Zealand is characterized by many more very small parties than is Germany, which we attributed to the incentives of both the alternative threshold (one district victory being sufficient to win proportional representation) and the attractiveness of entering cabinets or becoming a support partner to either major party. This latter feature may be subject to change in the future, as the "legacy politicians" who founded several of the smaller parties eventually retire and especially if the Elec- toral Commission's recommendation to abolish the alternative threshold is followed. In surveying the effects of several mixed-member systems, Shugart and Wattenberg (2001b) asked whether they could offer "the best of both worlds." While any such conclusions are necessarily normative, our overview of the consequences of MMP in New Zealand suggests that the answer may be yes. On the one hand, the system is overall proportional, as expected from its mechanical effect. On the other hand, voting in the single-seat districts has become relatively concentrated on the two main Political Consequences of New Zealand's MMP System 27 3 parties, and the presence of the nominal tier appears to allow for some degree of personal vote. In fact, legislators appear to maintain contact with local constituents (e.g., through electorate offices) even in districts where a different party has won the local contest. In these ways, we see the impact of both tiers shaping the political process in New Zealand. NOTES 1. Thus far the Left has not been considered a party with coalition potential. 2. In fact, the Green Party's roots go back before the Alliance was formed, and may be found in the Values Party that contested several elections (but won no seats) in the 1970s and 1980s. 3. While the term first past the post electoral system is more commonly used in New Zealand to describe the single-member plurality (SMP) electoral system, in this chapter we use the term SMP as part of the standardized usage of acronyms and terminology in this volume. 4. For a detailed typology of MMP and MMM systems, see Shugart and Wat- tenberg (2001a). 5. An "overhang" results when a party with less than 5% of the party-list vote obtains at least one district seat. Under current practice, not only does such a party qualify for list seats, but if its number of districts won exceeds its proportional share (of 120), the size of parliament is expanded to account for the overhangs. The Elec- toral Commission proposed ending both practices. In addition, it recommended retaining the status quo for by-election candidacy, dual candidacy (which we discuss below), and the ordering of party lists. 6. Nonetheless, the National Party, as of 2011, has actually formed three of the six post-MMP governments, albeit with some bargaining challenges, as we discuss in the section below on government formation. 7. As in New Zealand, there is a provision for overhang seats (iiberhangmandate). A key difference in Germany is that even though overall proportional entitlements to parties are determined nationwide, the overhangs are determined within each state. 8. As is typical, we count the Christian Democratic Union and the Christian Social Union as separate parties. The latter competes only in Bavaria, whereas the former competes in the rest of the country. The two act as a single party in parlia- ment. 9. If we were to count the CDU and CSU as one party, these figures would be about .84 as high, on average (3.97 in 2009). All but the 1990 figure (2.65) would remain higher than any value in New Zealand under MMP. 10. NS fell in 2013 in large part due to the Free Democratic Party failing to clear the threshold. 11. Counting CDU and CSU as one, the 1969-87 average is 2.44 and it would not be over 3.00 until 2005. 12. It is worth noting that, due to the difference in parliament sizes, Germany's alternative threshold of three districts won is not clearly higher than New Zea- land's one district. Germany's was equivalent to about 1.2% of total districts before 274 Mixed-Member Electoral Systems in Constitutional Context reunification but is about 0.5% of districts in the larger postunification Bundestag. New Zealand's is around 1.4% of districts. Nonetheless, it is presumably more dif- ficult for a very small party to find three candidates capable of winning districts- assuming it is not a party focused on a regional cleavage-than it is to have one such candidate. In the case of New Zealand, as we discuss later, that one candidate is often the party founder who already held the seat with one of the larger parties in the pre-MMP era. 13. In 1994, the Party of Democratic Socialism won three districts in the formerly Communist east, but only 4.4% of the nationwide party vote. In 1990, it won repre- sentation with only 2.4% of the nationwide party vote, but at that election there was a one-time variance in the threshold: a party had to win 5% in only one of the former separate countries of West Germany and East Germany, rather than nationwide. 14. In practice, a party that has one district win needs about 1.5% of the party vote to qualify for a second seat. 15. The Seat Product is the average district magnitude (M) multiplied by the size of the assembly (S). Taagepera's models predict that the number of seat-winning parties (of any size) is the fourth root of the Seat Product, while the effective num- ber is the sixth root: NS = (MS)1/6. 16. Actual parties are approximated by the inverse square root of the thresh- old (expressed in fractions rather than percent), and Ns by the inverse cube root, according to the logical models. 17. In Germany, given the greater difficulty of a small party winning despite being below the threshold, one would want to use only the threshold-based models. An estimated 4.47 parties (of any size) in parliament is close to the recently observed 5, but the observed effective number is now much higher than the estimated 2.71. Perhaps it is Germany, and not the newer MMP case of New Zealand, that is the hard case to explain! 18. The minimum in any one cabinet has been three if we count the two Christian Democratic parties (CDU and CSU) separately. While they always have entered government or gone into opposition as if they were one party, they maintain their separate party organizations and leadership. 19. This is in partial contrast to Japan and Taiwan, where preelectoral alliances have been more common. 20. Such agreements state that a minor party will not vote against the govern- ment in motions of confidence or supply, typically in exchange for some coopera- tion on policy areas of mutual interest. 21. As discussed in greater detail below, Maori voters under both the SMP and MMP systems have had the option of voting in separate single-member districts set aside to ensure representation of the Maori. The Royal Commission had recom- mended abolishing the Maori districts as unnecessary under a proportional system, but waiving the usual party-vote threshold for parties representing Maori. 22. A Labour defector, former finance minister Roger Douglas, was also a cofounder of ACT. 23. After a brief stint with a party called New Labour. 24. In the voter registration period, citizens are asked whether they claim Maori ethnicity and, if so, if they would like to be enrolled in the Maori roll. 25. See "Greens Given GM Warning," New Zealand Herald, March 6, 2002. The narrative here also draws heavily on Carroll (2005). Political Consequences of New Zealand's MMP System 27 5 26. "Labour Takes Green Road," New Zealand Herald, August 31, 2005; "Conces- sions Squeezed Out of Labour," New Zealand Herald, October 18, 2005. 27. "Greens Demand Two Posts in Cabinet," September 15, 2005. 28. "Labour-Led Government Co-operation Agreement with Greens," October 17, 2005 (accessed by authors from Green Party website on October 20, 2005). 29. In fact, the party went so far as to issue a press release on October 19, 2005 saying the presence of New Zealand First and United Future meant the govern- ment was "bad news for New Zealand workers." 30. "Labour Wins Crucial Votes on Emissions Trading Plan," New Zealand Her- ald, August 27, 2008. 31. The amendments to the ETS secured the support of the now National- aligned United Future. In addition, the Maori Party was pivotal, and for its support obtained a provision allowing carbon credits for the planting of trees on tribal land. See "Emissions Trading Money Scramble," The Press (Christchurch), November 23, 2009, and "'Sellout' Claim over ETS Deal," The Press (Christchurch), Novem- ber 24, 2009. The staunchly promarket ACT, also a National partner, voted against the amended ETS (preferring no ETS at all). 32. "Memorandum of Understanding between the New Zealand National Party and the Green Party of Aotearoa New Zealand," dated April 8, 2009 (accessed from the Green Party website in November 2009). 33. The Greens won an electorate (Coromandel) once, in 1999, thanks to some informal cooperation with the Labour Party, which did not withdraw its candidate from the district, but did encourage supporters to vote for the Green candidate in the otherwise National-leaning district. 34. In fact, the Commission's Report says that 3 % would be preferable and would not be so low as to threaten government stability. However, it concludes such a change would be too drastic and politically unpalatable, thereby settling on 4%. 35. Apparently the first use of the term contamination to refer to concurrent majoritarian and proportional balloting not being identical to the respective sys- tems when used separately is Shugart and Carey (1992, 239-42). They refer to cases of presidential elections using plurality and legislative elections using PR. In the literature on mixed-member systems, the concept of contamination appears to have been used first by Herron and Nishikawa (2001), who emphasized in a comparison of the Japanese and Russian MMM systems that the PR tier results in a lessened impact of the single-seat districts on the number of parties. 36. Even as early as the second MMP election, Karp et al. (2002) noted that split- voting followed a predictable pattern, with "strategic defections" from parties that were not viable in the voter's electorate. 37. The Scottish and Welsh MMP systems appear to be in between the effects we show here for Germany and New Zealand: a significantly lower Nv for the nominal tier compared to the list, but also average Nv in the districts that is greater than 3.0 (Crisp et al. 2012). 38. Horiuchi, Shugart, and Vowles (2013), analyzing the effect of redistricting on the vote in New Zealand, find evidence for a personal vote for incumbents, who tend to obtain fewer candidate votes from voters who were not in their district prior to redistricting. However, they do not find evidence for contamination, as voters who are redistricted are not less likely than other voters to vote for the list of the incumbent's party. 276 Mixed-Member Electoral Systems in Constitutional Context 39. Labour and National each have a few MPs who ran only on a list. Some of these were recruited from the private sector for their policy expertise (e.g., trade or health), while others represent minority ethnic communities. Some of these list- only candidates set up offices upon being elected, while others did not. 40. The two are Peters and the MP elected in Peters's former electorate of Tau- ranga, Brendan Horan. 41. See http://www.mmpreview.org.nz/proposals/dual-candidacy (last accessed February 27, 2013). The Commission acknowledges a submission to the MMP Review by Shugart as the source of this suggestion. REFERENC ES Bollard, Alan. 1994. "New Zealand." In The Political Economy of Policy Reform, ed. John Williamson. Washington, DC: Institute for International Economics. Boston, Jonathan. 2011. "Government Formation in New Zealand under MMP: Theory and Practice." Political Science 63 (1): 79-105. Carroll, Royce. 2005. "Legislative and Electoral Coalition Politics in MMP New Zealand, 1993-2003." Case study prepared for the Graduate School of Interna- tional Relations and Pacific Studies, University of California, San Diego. Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World's Electoral Systems. Cambridge: Cambridge University Press. Cox, Karen E., and Leonard J. Schoppa. 2002. "Interaction Effects in Mixed- Member Electoral Systems: Theory and Evidence from Germany, Japan, and Italy." Comparative Political Studies 35 (9): 1027-53. Crisp, Brian E, Joshua D. Potter, and John J. W. Lee. 2012. "Entry and Coordina- tion in Mixed-Member Systems: A Controlled Comparison Testing the Con- tamination Hypothesis." Journal of Politics 74 (2): 571-83. Denemark, David. 2001. "Choosing MMP in New Zealand: Explaining the 1993 Electoral Reform." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg. Oxford: Oxford Uni- versity Press. Fisher, Steven L. 1973. "The Wasted Vote Thesis: West German Evidence." Com- parative Politics 5:293-99. Gschwend, Thomas, Ron Johnston, and Charles Pattie. 2003. "Split-Ticket Pat- terns in Mixed-Member Proportional Systems: Estimates and Analyses of Their Spatial Variation at the German Federal Election, 1998." British journal of Political Science 3 3:109-27. Herron, Erik S., and Misa Nishikawa. 2001. "Contamination Effects and the Num- ber of Parties in Mixed-Superposition Electoral Systems." Electoral Studies 20 (1): 63-86. Horiuchi, Yusaku, Matthew S. Shugart, and Jack Vowles. 2013. "Electoral Redis- tricting and the Personal Vote under the Mixed-Member Electoral System: New Evidence from New Zealand." Paper presented at the 2013 Annual Meet- ing of the Midwest Political Science Association, Chicago, April 11-14. Karp, Jeffrey A. 2009. "Candidate Effects and Spill-over in Mixed Systems: Evi- dence from New Zealand." Electoral Studies 28:41-50. Karp,Jeffrey A., Jack Vowles, Susan A. Banducci, and Todd Donovan. 2002. "Strate- Political Consequences of New Zealand's MMP System 277 gic Voting, Party Activity, and Candidate Effects: Testing Explanations for Split Voting in New Zealand's New Mixed System." Electoral Studies 21:1-22. Lijphart, Arend. 1984. Democracies: Patterns of Majoritarian and Consensus Govern- ment in Twenty-One Countries. New Haven: Yale University Press. Lijphart, Arend. 1999. Patterns of Democracy. New Haven: Yale University Press. Miller, Raymond, and Jennifer Curtin. 2011. "Counting the Costs of Coalition: The Case of New Zealand's Small Parties." Political Science 63 (1): 106-25. Moser, Robert G., and Ethan Scheiner. 2012. Electoral Systems and Political Context: How the Effects of Rules Vary across New and Established Democracies. Cambridge: Cambridge University Press. Nagel, Jack H. 1998. "Social Choice in a Pluralitarian Democracy: The Politics of Market Liberalization in New Zealand." British journal of Political Science 28 (2): 223-67. New Zealand Electoral Commission. 2012. Report of the Electoral Commission on the Review of the MMP Electoral System. Report provided to the Minister of Justice for presentation to Parliament in accordance with section 78 of the Electoral Referendum Act 2010, Wellington, New Zealand, 29 October. http://www.elec tions.org.nz/sites/default/files/bulk-upload/documents/Final_Report_2012 Review_ofMMP.pdf, (accessed 10 August 2015). Nielson, Daniel L. N.d."The Politics of Economic Reform in New Zealand." Case study prepared for the Graduate School of International Relations and Pacific Studies, University of California, San Diego. O'Brien, Thomas. 2013. "Leaving the Minors: The Green Party of Aotearoa New Zealand and the 2011 General Election." Representation 49 (1): 69-81. Shugart, Matthew S. 2005. "Comparative Electoral Systems Research: The Matu- ration of a Field and New Challenges Ahead." In The Politics of Electoral Systems, ed. Michael Gallagher and Paul Mitchell. Oxford: Oxford University Press. Shugart, Matthew S., and John M. Carey. 1992. Presidents and Assemblies: Constitu- tional Design and Electoral Dynamics. Cambridge: Cambridge University Press. Shugart, Matthew S., and Martin P. Wattenberg. 2001a. "Mixed-Member Electoral Systems: A Definition and Typology." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg. Oxford: Oxford University Press Shugart, Matthew S., and Martin P. Wattenberg. 2001b. "Are Mixed-Member Sys- tems the Best of Both Worlds?" In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg. Oxford: Oxford University Press. Taagepera, Rein. 2007. Predicting Party Sizes: The Logic of Simple Electoral Systems. Oxford: Oxford University Press. Taagepera, Rein. Nd. "Predicting Party Sizes on the Basis of Legal Thresholds." Unpublished MS, Dept. of Political Science, University of California, Irvine. Taagepera, Rein, and Matthew S. Shugart. 1991. Seats and Votes. New Haven: Yale University Press. Vowles, Jack. 2012. "Legislative Accountability under New Zealand's MMP: Turn- over, Dual Candidacy, and 'Party-Hopping'." http://www.jackvowlesdomain. co.uk/Accountability_MMP.pdf (last accessed April 23, 2013). TEN Presidents and Blank Votes in the Bolivian and Russian Mixed-Member Systems Nathan F Batto, Henry A. Kim, and Natalia Matukhno Mixed-member electoral systems confront voters with two choices: a list tier ballot to be cast on the basis of the party and a nominal tier ballot to be cast on the basis of the characteristics of the local candidates. Some voters do not make these two choices independently, but instead use information about the choices in one tier to help make the choice in the other tier. These sorts of "contamination" effects have been the subject of a great deal of research (Cox and Schoppa 2002; Herron and Nishikawa 2001; Nishikawa and Herron 2004; and Ferrara, Herron, and Nishikawa 2005). In the introduction to this volume, Batto and Cox argue that directly elected executives shape the legislative party system differently than indi- rectly elected executives. We build on this idea and argue that in presiden- tial systems, a third consideration-what the voters know of the presiden- tial candidates-may also shape the vote choice on the list and nominal tier ballots. Presidential elections typically dominate public discourse during an election campaign, and voters often know much more about the presi- dential candidates than the party platforms or their local district candi- dates. Many voters might project what they know of a presidential candi- date onto his party's list or district nominee. However, we argue that the presidential candidates do not affect list and district tier decisions equally. The connections between the presidential candidate and the party list are typically much closer and more apparent, and voters may consequently be 278 Presidents and Blank Votes in the Bolivian and Russian Systems 279 less willing to assume that the district nominee is simply a local avatar for the presidential candidate. This is especially apparent in new democracies with less institution- alized parties that do not penetrate deeply into society. When a party is simply a hastily constructed vehicle for a presidential candidate's national campaign, local organizations are often weak, underfunded, and impotent. Many district candidates for such parties are likely to be unknown political novices without a credible chance of winning. Even people who turn out to support the presidential candidate and are willing to back the national party list may balk at voting for a completely unknown quantity with little chance of winning on the district ballot. Some countries allow voters to vote "blank" or "against all," and we examine patterns in these blank votes to infer how presidential candidates affect voting in the list and nominal tiers. An explicit blank vote can be interpreted as a sign that none of the candidates in that tier have inspired the voter to vote, so the impetus to vote must have come from elsewhere.' We examine two elections that took place in very different political envi- ronments, Bolivia in 2002 and Russia in 2003, and find that blank votes were much more common in the nominal tier than in the list tier. More important, larger numbers of blank votes were correlated with better per- formance in the list tier by more loosely organized parties led by personally popular and charismatic presidential candidates. These results are consis- tent with the idea that presidential candidates had a larger contamination effect on the list vote than on the district vote, and the effect was particu- larly strong for presidential candidates of newer and less organized parties. Contamination Effects and Presidential Candidates The basis of the contamination effect is commonly thought to be psycho- logical. Herron and Nishikawa suggest that voters value consistency in their vote choices. That is, if voters observe the same choices available on both ballots, they are inclined to make the same vote choices in both tiers (2001, 68-69). Cox and Schoppa also argue that the physical presence of the district candidate campaigning locally and furnishing a "human face" to a race has a positive impact for a party. Such a presence helps mobilize what they term "either way voters," those without a strong psychological or habitual commitment to that party. Moreover, the presence of a district candidate can increase the party's list tier totals, even if that district can- didate stands no chance of capturing the nominal tier seat (2002, 1031). 280 Mixed-Member Electoral Systems in Constitutional Context Ferrara, Herron, and Nishikawa (2005, 67) make the argument explicitly informational by explaining the effect as an informational shortcut: vot- ers can judge the quality of all the candidates on the party list by simply looking at the quality of the particular candidate in their local district. In the presence of this contamination effect, even small parties without any prospect of winning the election in the nominal tier are encouraged to field candidates in as many districts as they can in order to improve their list tier performance, thus countering the Duvergerian incentive to abstain from participating in unwinnable races. These arguments all look at how the district race contaminates the party list race. The most compelling evidence comes from parliamentary democracies, such as Italy, New Zealand, Germany, and Japan. This is no coincidence. Parliamentary systems often lack a single, nationalized party politician with a separate electoral mandate from his or her legislative counterparts, who can furnish a distinct "face" for the rest of the party (Samuels and Shugart 2010). By the nature of the institution, voters in a parliamentary system are not asked to vote explicitly for the party's national standard bearer in an election. While parties may conscientiously attempt to draw voter attention to their leaders, the linkage is at best indirect in the absence of an ability to vote directly for their standard bearers by name. Furthermore, party leaders in a parliamentary system often are not the most electorally profitable choices to act as their parties' "human faces." Party leaders in parliamentary democracies have often risen through the ranks on the basis of their skills as backroom operators in the legislative process, rather than on their ability to arouse the masses outside the cham- ber (Samuels and Shugart 2010). In other words, the voters in a parlia- mentary system can sometimes be confronted with a set of rather dry or uninspiring national party faces to base their party list decision on. Presidential systems are qualitatively different, especially when the presidential and legislative elections are held concurrently. Presidential candidates provide a visible and recognizable face. Often, they are political outsiders who have built a strong, well-defined reputation of their own. Indeed, many successful presidential contenders, regardless of the coun- try, are likely to have risen in politics through their ability to mobilize supporters through their own actions, rather than their roles in the inner workings of parties (Samuels and Shugart 2010). Media coverage is usually dominated by the presidential campaign, and the messages and reputations of local candidates are often drowned out amid the focus on the executive race. Such focus on the personal appeals of the presidential candidates is arguably justified. After all, the person elected president will be the head of Presidents and Blank Votes in the Bolivian and Russian Systems 281 government, and parties have no mechanism like a vote of no confidence to ensure that presidents from their party actually follow the party line. Quite the contrary, presidents almost always become the de facto party leaders and often decide to lead their parties in new directions chosen according to their personal priorities. Thus, not only is the information about the presi- dential candidate more easily accessible than corresponding information about a party leader in a parliamentary system, voters are fully justified in adjusting their attitudes about a particular party based on that information. Particularly in U.S. politics, the electoral impact of the president on legislative elections has long been recognized in numerous studies of the so-called coattail effect and its impact on legislative elections (Burnham 1975; Calvert and Ferejohn 1983; Campbell 1997; Miller 1955). However, the importance of the president may be even greater in other presidential democracies, especially where presidents enjoy greater formal power rela- tive to other political actors or where political parties have shorter histories and weaker reputations. Newer and less institutionalized parties are much more likely to field a reasonably charismatic presidential candidate than to be able to recruit a full slate of well-known candidates to run for the legis- lature. Modern media technology can help project the faces of presidential candidates to the homes of voters in every district, regardless of the efficacy of the local party organizations in promoting voters' familiarity with can- didates for the legislature. In short, information from the national race, or more accurately, the presidential race, is likely to be much stronger, clearer, and more omnipresent than information garnered from the local races. Presidential elections also affect turnout. More people turn out to vote in legislative elections when a presidential election is being held concur- rently. Perhaps more significantly, voters in concurrent elections also turn out to vote for different reasons.2 Since they have been bombarded with information about the presidential race, they are often thinking primarily about the presidential race and supporting a particular presidential can- didate. The legislative race is usually a secondary consideration. Indeed, many people may cast their legislative vote as an expression of support for the presidential candidate. If the presidential candidate is the human face of the party, the extent to which this human face spills over to the list and district votes may differ. The presidential candidate is usually the leader of the party. Even when he stresses positions that are at odds with orthodox party doctrine, other party leaders tend to downplay these differences since highlighting intra- party divisions during a campaign is usually a losing strategy. Hence, equat- ing a vote for the party list with support for the presidential candidate is 282 Mixed-Member Electoral Systems in Constitutional Context relatively straightforward. Doing so for the district candidate, however, is more complicated. District races are run along different contours from the national campaign. Districts may diverge significantly from the national average on many dimensions, such as ethnic makeup, partisan balance, and economic vitality. Different issues may matter locally, and local preferences may be at odds with national preferences. Because of this, a local party nominee may not always follow the national party in lockstep. Voters in local districts must also consider the question of viability. Party list seats are distributed proportionally, and even parties with relatively low levels of support can win list seats. This is not the case in the nominal tier with single-seat districts. Disorganized, underfunded, and unknown candidates are usually sure losers. Strategic voters may be loath to waste their district vote on such a candidate, even if they support the party's presidential can- didate. Alternatively, they may simply decide that voting for the party's district nominee is hopeless and choose not to vote at all in the nominal tier.3 In short, some voters will base their legislative votes on their opinions of the presidential candidate, and this is more likely to occur in the list tier than in the nominal tier. Blank Votes and Inchoate Parties In a number of countries, including Bolivia and Russia, voters have the option to vote "blank" or "against all"-to indicate that the voter does not want to support any of the candidates or parties. Blank votes are not the same as null votes, in which a voter votes for multiple candidates, does not mark any option, writes something inappropriate on the ballot, or otherwise spoils the ballot, and blank votes are counted separately from null votes.4 Why would a voter cast a blank vote? It is possible that the voter is highly informed about each candidate and simply dislikes all of them. However, most highly informed voters can eventually conclude that one candidate is a little better than the others, or at least that one is awful enough that it is worth it to vote for someone else in order to stop the worst candidate. Another possibility, which we suspect is more common, is that most blank votes reflect extremely low levels of information. That is, the typical blank vote occurs when the voter does not know enough about any of the options to decide that any of them are better (or worse) than the others. This begs the question of why such a voter would turn out in the first place. Of course, it could just be that the voter has a strong sense that voting is a democratic duty (Achen and Hur 2011) and believes that the Presidents and Blank Votes in the Bolivian and Russian Systems 283 actual choice is less important than the mere act of participation. More significantly for our purposes, a voter may turn out to vote because she has a strong opinion in a different race taking place at the same time. For example, with concurrent party list and district elections, the voter may have a strong opinion in the former elections but not in the latter race. In this case, the voter might vote for a party in the list tier and vote blank in the district tier. Since information about presidential candidates is so prevalent, and this information leaks primarily into the list tier and only secondarily into the district tier, most voters should be able to make a choice in the list tier. Information from the presidential race is less useful in local races, and more voters might be unable to differentiate among the various names on the ballot. Some of these voters will respond by explicitly not making any choice, by casting a "blank" ballot. Alternatively, voters who turn out expressly to support the presidential candidate may vote for the party list, which is closely associated with the president, but not the district candi- date, who may not be viable or may take somewhat different positions. More of these voters should vote blank in the nominal tier than in the list tier. This implies: H1: In countries with a presidential system and a mixed-member electoral system, blank votes should be more common in the dis- trict tier than in the list tier. Hi is fairly crude; it makes only a single, national-level prediction. How- ever, we can use the differences between more institutionalized and less institutionalized parties to draw a more finely tuned hypothesis. The rela- tive degree to which the presidential candidate defines the "face of the party" varies from party to party. We are particularly interested in parties that revolve around a single leader. Where these types of parties are strong, blank votes should be more common. These personal vehicle parties are often hastily organized for the upcoming presidential campaign, and the rest of the party is somewhat of an afterthought. Ironically, it may be precisely the presidential candidates who most thoroughly embody their party image who have the most trou- ble extending that image and influence voting down to the local level. It is often much harder for new and disorganized parties to assemble a roster of competent local candidates. Instead, these inchoate parties are often only able to cobble together a group of unknown, unimpressive, and under- funded local politicians. Such local politicians will have difficulty commu- 284 Mixed-Member Electoral Systems in Constitutional Context nicating information about themselves to the voters, so they cannot com- pete with, modify, or reshape the image projected by the national party. However, they might also be so anonymous that there is effectively no local presence for the national party face to be projected onto. The more invisible the local candidate, the more likely voters are to decide that there is no viable foot soldier for their favored presidential candidate. Almost by definition, an anonymous candidate is an incompetent candidate. In such a district, voters might decide simply to vote for the presidential candidate's party in the list tier and to vote blank in the nominal tier. Older, more established, or less personalized parties are likely to have a stable of fairly well-known local candidates supported by a substantial net- work of organizations at the grassroots level. The better the organization, the easier it is for politicians to familiarize and ingratiate themselves with their voters. Access to such organizational resources permits candidates of more established parties a greater opportunity than their counterparts in more inchoate parties to communicate to the voters and, in so doing, assert their viability. In such a party, the presidential candidate is not the sole face of the party, and local candidates may even try to present a slightly different platform or image. However, because of the national focus on the presidential race, the presidential candidate will still be the dominant face of the party. Even the strongest local candidates are generally happy to cloak themselves in the aura of their party's presidential candidate. Vot- ers in these situations who want to support the presidential candidate have a viable candidate and a clear strategy. In essence, the difference between more and less organized parties is that the roster of viable candidates fielded by the former has the capacity to soak up all the votes from sup- porters of the presidential candidate while the roster of the latter may not. Many parties in presidential systems are formed precisely as personal vehicles to promote the political fortunes of a politician with presidential ambitions.5 These parties are often hastily organized and lack any sort of organizational capacity to penetrate deeply into society. Indeed, the rest of the party is somewhat of an afterthought. For these parties, the pro- jection from the presidential candidate to the party list should be quite strong. Since the presidential candidate dominates the party, the top spots on the party list usually go to his close allies or cronies, and, with a lower electoral threshold, viability is less of a problem. However, parties formed as personal vehicles are precisely the types of parties that find it difficult to cobble together a full roster of competent local candidates.6 They may have strong local candidates in some districts, but these newer parties typi- cally round out the roster with unqualified, unknown, underfunded, or oth- Presidents and Blank Votes in the Bolivian and Russian Systems 285 erwise dismal candidates. At the extreme, they may not even be able to field a candidate at all. Such local politicians will have difficulty communicating information about themselves to the voters, so they cannot compete with, modify, or reshape the image projected by the national party. However, they might also be so anonymous that there is effectively no local presence for the national party face to be projected onto. The more invisible the local candidate, the more likely voters are to decide that there is no viable foot soldier for their favored presidential candidate. Ironically, it may be precisely the presidential candidates who most thoroughly embody their party image who have the most trouble extending that image and influence voting down to the local level. In short, whereas the rosters of established parties tend to be filled with competent local candidates, inchoate parties have a higher percentage of incompetent local candidates. Since voters are more likely to vote blank in the nominal tier when the local candidate of their favored party is incom- petent or missing altogether, parties that are newer, less organized, or per- sonal vehicles should be more associated with blank votes in the nominal tier than older and more institutionalized parties. This leads to our second hypothesis: H2: In countries with a presidential system and a mixed-member electoral system, blank votes in the nominal tier should be more common in areas in which less institutionalized parties win a higher vote share in the list tier. Methodology and Case Selection Methodologically, this paper employs a most different systems research design. In this research design, cases that are different on a set of control variables are compared. However, the independent variables of interest and the dependent variables of the cases should be the same. If the theorized process works the same way, with the same independent variables leading to identical outcomes, in very different contexts, this provides confidence that the outcome is not the result of some third variable (Przeworski and Teune 1970, chap. 2). Given the preceding theoretical discussion and hypothesis, we need cases with directly elected presidents, mixed-member electoral systems for the legislature, the option to vote blank, and a party system in which par- ties' capacity to field a roster of credible local candidates varies. The cases 286 Mixed-Member Electoral Systems in Constitutional Context should be different in as many other aspects as possible. The patterns of blank votes should then vary in predictable and similar ways. We have chosen to examine Bolivia in 2002 and Russia in 2003. Both countries have a presidential system and elect their national legislature by a mixed-member system. In Bolivia, voters can vote "blank," while in Rus- sia, voters can cast a vote "against all." The party systems are discussed below, and we argue that both cases featured some parties that could pres- ent credible candidates in most races and other parties that did not have such broad-based organizations. In the remainder of this section, we note some of the more obvious differences between the two cases. Russia is much larger, ecologically diverse, and more populous. Bolivia was a Spanish colony, while Russia was the heart of an empire. In the Cold War, Russia was a central player while Bolivia was a peripheral pawn. Following the Russian Revolution of 1917, most property was collectivized and controlled by the state. After the Bolivian Revolution of 1952, land was seized from landlords and given to local communities and syndicates who held it collectively on behalf of the local peasants (Klein 2003, 214-15). Bolivia remains a much more rural and agricultural society. Ethnically, both countries are diverse, but Russia is dominated by ethnic Russians, while Bolivia has historically been domi- nated by its white minority rather than by the indigenous peoples who collectively make up a majority.7 In recent years, indigenous peoples have transformed their numbers into political power through extensive grass- roots organizing by social organizations. There has been no equivalent wave of grassroots social movements in Russia. In Bolivia, the mobiliza- tion of large numbers of previously inactive citizens led to an enormous increase in voting.8 By contrast, turnout has declined in Russia.9 There are a few notable differences in the two electoral systems. First, Russia had an MMN system, while Bolivia employs an MMP system. Sec- ond, Bolivia uses a fused vote. Voters cast two ballots, and the nominal tier ballot elects the local district seat for the lower house. The second ballot is more consequential. It not only decides the seat share for each party in each department, it is also used to elect the president and the Senate. In fact, what we are calling the list tier ballot is better understood as a presidential ballot that is also applied to legislative seats. In this chapter, we assume that presidential candidates effectively become the face of their party. The Bolivian fused vote almost necessarily assures that this is the case, since a vote for the presidential candidate is indistinguishable from one for the party list. However, Russia does not use a fused vote. In 2003, voters cast three separate ballots for the district, party list, and president. A Presidents and Blank Votes in the Bolivian and Russian Systems 287 third difference is that parties in Bolivia run full slates of candidates. One possible reason for a blank vote is that the voter's favorite party may not run a candidate in the local district race. Rather than vote for another, less favored party, the voter might simply vote blank. This was certainly a pos- sibility in Russia, but it probably does not drive the overall results because Bolivian parties present full slates of candidates. That is, every party that runs one candidate for president or the legislature runs a candidate for every seat available. In 2002, 11 parties presented presidential candidates, each of the 68 legislative districts had 11 candidates, and each of the e11 parties listed 62 candidates for the 62 party list seats. Unaffiliated candi- dates were not allowed, so the partisan choice set was exactly the same for every voter. If a Bolivian voter had a favorite party, she could vote for that party on both ballots. The Bolivian Party System in 2002 In the later part of the 2000s, the Bolivian party system underwent a dra- matic change, evolving into a dominant party system. However, in 2002 the Movimiento al Socialismo (MAS) was just beginning its rise to power, and it was just one of several contenders for power. The party system was a rather fragmented multiparty system, and no party had won an absolute majority in votes or seats since the restoration of democracy in the early 1980s. Three parties, the Movimiento Nacionalista Revolucionario (MNR), the Acci6n Democratica y Nacionalista (ADN), and the Movimiento de Izquierda Revolucionaria (MIR), dominated politics throughout the 1980s and 1990s. All the presidents in that era came from one of these three par- ties, and they collectively won over half the vote in every election. The MNR and MIR can both trace their roots to the 1940s, while the ADN was formed in the 1970s by the former military dictator Hugo Banzer (Klein 2003). Ideologically, the MIR presents itself as a leftist party, the MNR as a centrist party, and the ADN as a rightist party. However, all three signed on to the neoliberal economic reforms of the 1980s and 1990s, and by 2002 there was not a great difference among the three traditional parties' stances on economic issues (Singer and Morrison 2004, 174). Instead, the three parties' primary focus was on distributing patronage to their follow- ers (Gamarra and Malloy 1995; Domingo 2005, 1731). Prior to the electoral reform of 1994, seats in the lower house of the legislature were elected by closed list PR. As might be expected, power within the traditional parties was characterized as centralized and hierar- 288 Mixed-Member Electoral Systems in Constitutional Context chical. In particular, nominations were tightly controlled by the top lead- ers (Gamarra and Malloy 1995, 419). Indeed, one of the main reasons for the electoral reform was to encourage closer ties between politicians and their voters (Mayorga 2001a, 201), and stronger pressures for district can- didates to develop local reputations began to emerge after reform (May- orga 2001b, 438-42). At the same time, the Law of Popular Participation passed in 1994 marked a significant devolution in political power from the central government to localities, making it much easier for local politicians to establish local power bases (Van Cott 2008). By 2002, the three tradi- tional parties had been concentrating on distributing patronage for over a decade. Much of this patronage went through local politicians who used it to build personal reputations.'0 As such, by 2002 the three traditional parties could field rosters of candidates who would have been quite well known in their local districts. A second group of parties was quite different. In 2002, two new parties emerged as contenders for power, finishing second and third in the presi- dential election. The Nueva Fuerza Republicana (NFR) was led by Man- fred Reyes Villa, a former prefect of Cochabamba. The NFR was a vehicle for Reyes Villas's presidential candidacy, and it declined quickly after his defeat in the 2002 election. The Movimiento al Socialismo won second place over the NFR by fewer than a thousand votes. The MAS was dominated by a campesino movement headed by Evo Morales. Morales had been elected to the leg- islature in 1997 under the banner of the Izquierda Unida, but had split off to form his own organization in 1998. Because the costs of establishing a formal political party are high, MAS agreed to let Morales and his organi- zation run under their banner in 2002 (Van Cott 2008, 53). While Morales has now been the dominant force within MAS for a decade, the party in 2002 was less a coalition of politicians than an umbrella group of heteroge- neous social movements supporting a charismatic leader (Mayorga 2008). While MAS had candidates with local organizations and popularity in some areas, this network was patchy. Moreover, the leaders of these move- ments were primarily social activists and only secondarily candidates for political office. These two very different challenges require very different types of organizations, tactics, and appeals. A third party, the Movimiento Indigena Pachakuti (MIP) vied with the MAS for leadership of the movement for indigenous people's rights, tak- ing a more aggressively proindigenous stance (Domingo 2005, 1737-38; Madrid 2005, 165). Like the NFR and MAS, it was dominated by a char- Presidents and Blank Votes in the Bolivian and Russian Systems 289 ismatic leader, Felipe Quispe. Van Cott suggests that "the MIP is more a personalist vehicle for Quispe than the electoral expression of a social- movement organization. It is a classic charismatic party in that the leader determines party goals, chooses all militants and candidates, mediates all disputes, oversees all communications, distributes all incentives, and pre- vents the institutionalization of structures or leadership beyond his per- sonal control" (2008, 55). The NFR, MAS, and MIP looked very different from the three tradi- tional parties in 2002. Whereas the latter had networks of established local politicians, the former were dominated by charismatic leaders. Their ros- ters of district candidates were almost certainly characterized by candidates who had not been funneling patronage resources into building personal followings for several years. In the 2002 election, the three traditional parties won 42% of the list vote, while the three new parties won 48%. Five small parties combined to win the remaining 10%. In this paper, we will not consider the five small parties. The Russian Party System in 2003 Not unlike the Bolivian elections in 2002, the Russian elections in 2003 featured a fragmented mixture of parties, none of which enjoyed anything approaching dominant electoral support. Most of the parties were rela- tively new, having been born in the chaotic 1990s following the collapse of the Soviet Union. Indeed, only one, the Communist Party of the Rus- sian Federation (KPRF), the successor to the Communist Party of Soviet Union, could be considered an "old" established party among the con- tenders in 2003. The newer parties differed among themselves in terms of the paths to party organization they adopted. Some were built around centralized leadership of charismatic leaders without much organization at the local level while others established close connections with the central government in Moscow and leveraged their access to power to build orga- nizational networks that could support local candidates. The KPRF inherited substantial organizational assets from the old Communist Party of the Soviet Union. In addition, many of its mem- bers retained control over administrative resources in much of the coun- try throughout 1990s, especially in the southern regions of Russia, which became known as the "red belt" on the account of continued strong sup- 290 Mixed-Member Electoral Systems in Constitutional Context port that the party retained (Hadenius 2002). Its leader, Gennady Zuganov, was a former bureaucrat famously lacking in charisma and with little per- sonal following (White 2005). United Russia, the party associated with Vladimir Putin, would soon become the dominant party, but in 2003 it was still merely a plurality party. United Russia had been constructed on the foundation of earlier attempts to build a party of power (Gel'man 2008). In the mid-1990s, Yeltsin sup- porters had formed Our Home Russia as a party of power to support the Kremlin in the Duma. In 1999, these Kremlin insiders were challenged by a coalition of local governors operating under the label of Fatherland, and Fatherland almost won a majority in the Duma. In an effort to consolidate power and tie the powerful local governors to the Kremlin, Our Home Russia merged with Fatherland to create a new party, called Unity. This new party, which was later renamed United Russia, was meant to promote penetration into Russian society and prevent future strong challengers from emerging (Hale 2006). United Russia relied heavily on local gov- ernors, other officials supportive of the Putin administration, and on the country's largest companies (White 2005). Each brought their respective constituencies into the coalition as well as impressive abilities to mobilize the electorate. The Union of Right Forces (URF) was founded by high-level officials in the Yeltsin government who had been associated with attempts to imple- ment liberal economic reforms and continued to push for a deepening of market liberalization and democratization (Kara-Murza 2001). The URF envisioned itself as a party that would work within the power structure whenever possible, cooperating with the authorities to implement their preferred policies (Klyamkin 2004), and it continued to enjoy some sup- port from and access to the Kremlin during the early 2000s (Sakwa 2005; Gel'man 2008). United Russia and the URF thus constituted the de facto legislative coalition supporting the Kremlin and its program in the early years of Putin's administration (Rose, Munro, and White 2001; Sakwa 2005). This connection to the regime likely yielded them the resources necessary to build organization at the local level, to recruit a superior slate of local can- didates, and certainly to better familiarize voters with their candidates. Three other parties, the Liberal Democratic Party of Russia (LDPR), Yabloko, and Motherland, had far less organizational strength or local reach. The LDPR serves as a personal vehicle for its charismatic leader, Vladimir Zhirinovsky. Zhirinovsky's appeal centers on economic issues and nationalism. LDPR attempted to create strong grassroots organiza- Presidents and Blank Votes in the Bolivian and Russian Systems 291 tions during the 1990s, but these efforts have largely failed. While it still maintains regional branches, caucuses, and committees, its regional leaders remain mostly unknown (Hadenius 2002). Similarly, Yabloko was dominated by a charismatic leader, Grigory Yavlinsky. Yabloko tried to model itself along the lines of European social democratic parties (White 2005), but its electoral and organizational bases remained confined to Moscow and St. Petersburg (Hadenius 2002). Unlike the other major liberal force, the URF, Yabloko operated as a pure opposi- tion party, systematically opposing government proposals, and it did not enjoy access to a stream of resources from the Kremlin (Gel'man 2008). The third party of this group, Motherland, was only created a few months before the 2003 elections and had very little time to develop an organization or establish a reputation. It was founded by federal and local elites who were opposed to Putin and United Russia. Motherland hoped to siphon away supporters from the KPRF and LDPR, and the party appealed to nationalist voters who supported some socialist ideals but were dissatis- fied with the Communists (White 2005). In sum, we identify the KPRF, United Russia, and the URF as more institutionalized parties at the time of 2003 election, and the LDPR, Yabloko, and Motherland are classified as less institutionalized. Data and Results H1 makes a simple national-level prediction, that there should be more blank votes in the nominal tier than in the list tier. This expectation is clearly realized. Table 10.1 summarizes the results of the two elections. In Bolivia in 2002, 12.0% of the votes in the district tier were blank, nearly three times the 4.4% blank votes in the list tier. In Russia in 2003, the dis- parity was similar. Only 4.7% of the votes in the list tier were blank, but 13.3% of voters cast blank votes in the district tier. Note that this pattern is not confined to the two elections we are studying. Table 10.2 shows that there have been far more blank votes in the district tier than in the list tier in all eight elections held under mixed-member rules. In fact, the trend was particularly dramatic in Bolivia in 2005 and 2009. In 2005, a whop- ping 25.2% of district votes were blank, while only 4.0% of list votes were. In 2009, there was again an enormous disparity, with 21.4% blank in the district tier and 3.3% blank in the list tier. The consistently high numbers of blank votes in the districts tier is a stunning finding, especially from the perspective of mobilization. One of the old tropes of political campaigns TABLE 10.1. Summary of 2002 Bolivian and 2003 Russian Elections List tier District tier votes % votes % Bolivia 2002 MNR (M) 624,126 22.5 611,027 23.9 MAS (L) 581,884 20.9 373,454 14.6 NFR (L) 581,163 20.9 395,559 15.5 MIR (M) 453,375 16.3 478,614 18.7 MIP (L) 169,239 6.1 134,177 5.3 ADN (M) 94,386 3.4 190,618 7.5 Other parties 274,635 9.9 371,670 14.5 Valid votes 2,778,808 100.0 2,555,020 100.0 Null 84,572 2.8 73,417 2.5 Blank 130,685 4.4 359,380 12.0 Total votes 2,994,065 100.0 2,987,817 100.0 Russia 2003 United Russia (M) 22,776,294 40.1 14,124,122 28.6 KPRF (M) 7,647,820 13.5 6,119,161 12.4 LDPR (L) 6,944,322 12.2 1,827,716 3.7 Motherland (L) 5,470,429 9.6 1,719,147 3.5 Yabloko (L) 2,610,087 4.6 1,54,1463 3.1 Union of Right 2,408,535 4.2 1,757,146 3.6 Forces (M) Other Parties 8,975,297 15.8 7,173,919 14.5 Independents 15,127,780 30.6 Valid votes 56,832,784 100.0 49,390,454 100.0 Null 948,435 1.6 1,247,512 2.1 Blank 2,851,958 4.7 7,745,248 13.3 Total votes 60,633,177 100.0 58,383,214 100.0 Notes: (M) indicates party is classified as more institutionalized; (L) indicates party is classified as less institutionalized. Vote shares for parties and candidates are calculated using valid votes; shares of null and blank votes are calculated using total votes. TABLE 10.2. Blank Votes in Mixed-Member Elections in Bolivia and Russia List tier District tier Bolivia 1997 3.3% 8.7% 2002 4.4% 12.0% 2005 4.0% 25.2% 2009 3.3% 21.4% Russia 1993 4.2% 15.4% 1995 2.8% 9.9% 1999 3.3% 11.9% 2003 4.7% 13.3% Source: All Bolivian data and the Russian data for 2003 are from the official election commissions. Russian data for 1993- 99 come from the Project for Political Transformation and the Electoral Process in Post-Communist Europe dataset (http:// www.essex.ac.uk/elections/). Presidents and Blank Votes in the Bolivian and Russian Systems 293 is that you just have to get your voters to the polls. However, in most of these elections, one-tenth of the voters who were at the polls chose not to vote for any district candidates. One imagines throngs of voters coming in to the polls, voting for the president, then turning to a district ballot full of bewilderingly anonymous names, and either voting for the candidate with the same party label as their preferred presidential candidate or simply vot- ing blank to fulfill their obligations as citizens to participate. H2 states that there should be more blank votes in the district tier when a new and less organized party does better in the list tier. Based on the discussion in the previous two sections, we can crudely classify the various parties as more or less institutionalized. For each country we ran two simple OLS regressions. We analyzed the 2002 Bolivian and 2003 Russian elections using sub-district-level election returns. In Bolivia, we examined the 19,876 precincts, while in Russia we looked at the 2,756 towns. The dependent variable is the excess of blank votes in the district tier. This is calculated as the percentage of blank votes in the district tier minus the percentage of blank votes in the list tier." The independent variables are the party list vote shares of the more institution- alized and less institutionalized parties. We expect the coefficient for the less institutionalized parties to be larger than that for more institutional- ized parties, reflecting a greater number of blank votes where less institu- tionalized parties are more popular. We operationalize the independent variables in two different ways. The simpler method is to add all the less (more) organized parties together to see the overall pattern. We also exam- ine the individual parties to make sure that the pattern applies to each party within the group. The results are in table 10.3. Models 1 and 3 compare the two groups of parties. The coefficients for the more organized parties should be smaller than those for the less organized parties.'2 In both models, this is, in fact, the case. That is, higher numbers of blank votes in the district tier (control- ling for blank votes in the list tier) are associated with better performances by less organized parties in the list tier. To illustrate the magnitude of this effect, we might consider some hypothetical cases. In Bolivia, consider a precinct with 45% for the more organized parties in the list tier, 45% for the less organized parties, and 10% for all other parties. In Russia, consider a town with 60% for the more organized parties, 25% for the less orga- nized parties, and 15% for all others. These cases, which are fairly close to the national averages, should yield 7.2% and 8.1% more blank votes in the district tier than in the list tier, respectively. If we make a modest change by giving the organized parties 10% more and the less organized parties 10% 294 Mixed-Member Electoral Systems in Constitutional Context less, the excess blank votes decline to 6.1% and 4.9% in the Bolivian and Russian examples, respectively. Even relatively small changes are sufficient to produce notable differences in the percentages of blank votes. In Models 2 and 4, we disaggregate the groups into individual parties in order to ensure that these results apply to each individual party. In fact, we find exactly the patterns we expect. In both countries, all three of the less organized parties have larger coefficients than any of the more organized parties. Moreover, the differences between these coefficients are statisti- cally significant. Table 10.4 looks at post-hoc pairwise comparisons of the coefficients. In Bolivia, the coefficients for all parties are significantly dif- ferent at the p < .001 level. In Russia, there are pairs within each group that are not significantly different, but all pairs between groups are significant. Overall, we find clear support for our hypotheses. There are far more blank votes in the district tiers than in the list tiers. Moreover, higher per- centages of district blank votes are associated with better list tier perfor- mance by relatively inchoate parties. TABLE 10.3. OLS Regression of Blank Votes and Party List Vote Shares b s.e. t Bolivia 2002 Model 1 Constant .120 .006 19.84 Adj R2 = .156 Less organized .002 .007 .31 N= 19,876 More organized -.109 .007 -16.00 Model 2 Constant .127 .006 21.06 Adj R2 =.184 MIP .052 .007 7.36 N= 19,876 MAS -.028 .007 -4.17 NFR -.009 .007 -1.20 MIR -.084 .007 -11.36 MNR -.121 .007 -16.77 ADN -.182 .009 -19.43 Russia 2003 Model 3 Constant -.053 .012 -4.46 Adj R2 = .254 Less organized .381 .017 22.27 N= 2,757 More organized .064 .014 4.52 Model4 Constant -.014 .013 -1.10 Adj R2 = .277 Yabloko .599 .049 12.22 N= 2,757 LDPR .326 .027 11.92 Motherland .278 .027 10.37 United Russia .038 .015 2.63 URF -.006 .042 -0.06 KPRF -.065 .024 -2.74 Presidents and Blank Votes in the Bolivian and Russian Systems 295 Conclusion This volume examines many ways in which the constitutional system affects other elements of the political system. Various chapters examine the effects of the constitutional system on the party system, the faction system, the way candidates are nominated, the ways in which candidates campaign, and so on. This chapter examines another facet of this theme: how the electoral system, in conjunction with the presidential system, shapes the linkage between elections for different offices. In particular, presidential candidates shape the legislative elections by acting as the public faces of the parties that color voters' evaluations of legislative parties and candidates. Our perspective departs from the previous studies of contamination effects that have generally focused on parliamentary systems and have emphasized how local district candidates influence national party list choices. In contrast, this chapter argues that the presidential system pro- vides a basis for a potentially powerful contamination effect in the oppo- site direction. Voters project what they know of the presidential candidate onto the party list and, to a lesser extent, local district candidates. This complements, rather than repudiates, the existing research by elucidating a broader institutional context in which such intertier electoral interactions take place. By virtue of having the chief executive subject to a direct elec- tion, presidential systems encourage recruitment and development of char- ismatic party leaders who can successfully appeal directly to the broad elec- torate, with an independent political base of their own. The institutional TABLE 10.4. Pairwise Comparison of Regression Coefficients from Table 10.3 Bolivia MAS NFR MIR MNR ADN MIP * * * * * MAS * * * * NFR * * * MIR * * MNR * Russia LDPR Motherland UR URF KPRF Yabloko * * * * * LDPR * * * Motherland * * * United Russia * URF Note: An asterisk (*) indicates the difference between the two coefficients is statisti- cally significant; the probability that they are equal is p < .0001. 296 Mixed-Member Electoral Systems in Constitutional Context role normally required of parliamentary systems, of appealing largely to a relatively limited group of party elites applies comparatively less to those in a presidential system. Leaders in a presidential system, thus, tend to be better equipped to exert an influence on the vote choices in legislative elections, even if the formal institutional linkages may be potentially more tenuous. This chapter, however, also observes that the extent to which the presi- dential election can contaminate the legislative election is constrained by the nature of the parties. Counterintuitively, the more a presidential can- didate monopolizes the party image, the less he may be able to project the party face onto local candidates. We have shown that the district can- didates representing less institutionalized parties do not appear to be as able to capitalize on the popularity of their own presidential candidates as their better institutionalized counterparts. In parties organized around a single presidential candidate and with anonymous local candidates, voters may not see the local nominee as an avatar for the presidential candidate. Rather, they might simply conclude that there is no viable local foot soldier for their favored presidential candidate. To successfully project the party face, the local candidate must meet a certain standard of competence. That is, presidents most successfully act as the face of their parties when the party can field a full roster of candidates who each have their own local identities, resources, and networks in society. Shared party label alone, then, is not a sufficient condition for the separation of mandates inherent in a presidential system to be overcome in the electoral arena. The institu- tional separation between elections to different offices remains significant: successfully bridging this gap requires a superior party organization at the local level. NOTES 1. We draw distinction between an explicit blank vote, where the voter deliber- ately chose no particular candidate, from an invalid vote, where the nonchoice of the voter may be due to technical reasons other than actual voter choice. 2. This linkage is important in the pattern of "Surge and Decline" well known from studies of U.S. presidential and off-year elections (see Jacobson 2009 for an overview). 3. The incentives to vote strategically differ in MMM and MMP. In MMM, standard Duvergerian logic applies. Voters have an incentive to avoid wasting their votes on lower ranking candidates. In MMP, the seat shares of each party are deter- mined by the list tier vote, so incentives to vote strategically are much weaker. After all, if a party loses a particular nominal seat, it will be compensated with one additional list seat. On the one hand, this frees voters to vote sincerely for their Presidents and Blank Votes in the Bolivian and Russian Systems 297 favorite party. On the other hand, if they do not have strong feelings about the local race, voters are also free to abstain or vote blank. While the logic differs, the critical point for this paper is that in both MA/MM and MMP there are reasons for voters who support a national party with an uncompetitive local candidate to eschew vot- ing for that local candidate. 4. In this paper, we consider both blank and null votes to be invalid votes. That is, they are included in calculating turnout, but they are not included when calculat- ing the vote share of candidates or parties. This is consistent with the practice in Bolivia, but not in Russia. In Russia, both blank and null votes are counted in the divisor for the purposes of determining whether a party has passed the 5% thresh- old to receive list tier seats. 5. For example, Hicken notes that in the Philippines the number of legislative parties is inflated by the number of presidential candidates. Many of the parties are newly formed personal vehicles of a particular presidential candidate and are typi- cally not highly institutionalized (Hicken, chapter 8, this volume). 6. It is not always the case that hastily organized parties cannot recruit a large number of competent local figures. For example, in Taiwan James Soong's People First Party was hastily organized before the 2001 legislative election and attracted many established politicians. We do not consider this sort of party as an incho- ate party. The key question is not whether the party has a charismatic leader, but whether the leader stands alone as nearly the only politician with a popular follow- ing in the party. 7. In the 2006 LAPOP survey, Bolivians self-identified as 64.8% mestizo, 19.3% indigenous, and 11.0% white. On a separate question asking which, if any, indige- nous group they belonged to, 36.4% said they were Quechua, 24.2% were Aymara, 10.9% mentioned another indigenous group, and 28.5% belonged to no indig- enous group (Seligson et al. 2006). In general, self-identification as indigenous has increased markedly since 1998. 8. In 1989, 1.57 million votes were cast in the presidential election; in 2002, the number of votes nearly doubled to 2.99 million. 9. Turnout in Russia was 54.2% in 1993, 64.3% in 1995, 61.3% in 1999, and 55.3% in 2003 (Nohlen and Stover 2010). 10. See Lazar 2004 for a case study of one such politician. 11. These percentages are calculated using total votes. Party list vote shares are calculated using valid votes only. 12. The constant is determined by the minor parties. These parties are outside the scope of this paper, so the fact that the constant is positive in Model 1 and nega- tive in Model 3 is unimportant. REFERENCES Achen, Christopher H., and Aram Hur. 2011. "Civic Duty and Voter Turnout in Japan and South Korea." Election Studies 1 (2): 45-69. Burnham, Walter Dean. 1975. "Insulation and Responsiveness in Congressional Elections." Political Science Quarterly 90:411-3 5. Calvert, Randall, and John Ferejohn. 1983. "Coattail Voting in Recent Presidential Elections." American Political Science Review 77 (2): 407-19. 298 Mixed-Member Electoral Systems in Constitutional Context Campbell, James E. 1997. The Presidential Pulse of Congressional Elections. Lexington: University of Kentucky Press. Cox, Karen, and Leonard Schoppa. 2002. "Interaction Effects in Mixed-Member Electoral Systems: Theory and Evidence from Germany, Japan, and Italy." Comparative Political Studies 3 5 (9): 1027-5 3. Domingo, Pilar. 2005. "Democracy and New Social Forces in Bolivia." Social Forces 83 (4): 1727-44. Ferrara, Federico, Erik Herron, and Misa Nishikawa. 2005. Mixed Electoral Systems: Contamination and Its Consequences. New York: Palgrave Macmillan. Gamarra, Eduardo A., and James M. Malloy. 1995. "The Patrimonial Dynamics of Party Politics in Bolivia." In Building Democratic Institutions: Party Systems in Latin America, ed. Scott Mainwaring and Timothy R. Scully. Stanford, CA: Stanford University Press. Gel'man, V. 2008. "Party Politics in Russia: From Competition to Hierarchy." Europe-Asia Studies 60 (6): 913-30. Hadenius, A. 2002. The Development of Political Parties: Russia in Perspective. Glasgow: Centre for the Study of Public Policy, University of Strathclyde. Hale, Henry E. 2006. Why Not Parties in Russia? Democracy, Federalism, and the State. Cambridge: Cambridge University Press. Herron, Erik, and Misa Nishikawa. 2001. "Contamination Effects and the Number of Parties in Mixed-Superposition Electoral Systems." Electoral Studies 20 (1): 63-86. Jacobson, Gary C. 2009. Politics of Congressional Elections. 7th ed. New York: Pearson- Longman. Kara-Murza, Alexey. 2001. "At the Intersection of Politics and Science" [in Rus- sian]. Virtual Esse-Polis-Political Studies (December). http://www.politstudies.ru/ universum/esse/index.htm#02_04. Accessed June 6, 2012. Klein, Herbert S. 2003. A Concise History of Bolivia. New York: Cambridge Univer- sity Press. Klyamkin, Igor. 2004. "Postmodernity within Traditionalist Dimensions" [in Rus- sian]. Polis-Political Studies (January). Lazar, Sian. 2004. "Personalist Politics, Clientalism, and Citizenship: Local Elec- tions in El Alto, Bolivia." Bulletin of Latin American Research 23 (3): 228-43. Madrid, Raul L. 2005. "Indigenous Parties and Democracy in Latin America." Latin American Politics and Society 47 (4): 161-79. Mayorga, Rene Antonio. 2001a. "Electoral Reform in Bolivia: Origins of the Mixed-Member Proportional System." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg. New York: Oxford University Press. Mayorga, Rene Antonio. 2001b. "The Mixed-Member Proportional System and Its Consequences in Bolivia." In Mixed-Member Electoral Systems: The Best of Both Worlds?, ed. Matthew Soberg Shugart and Martin P. Wattenberg. New York: Oxford. Mayorga, Rene Antonio. 2008. "Populism in Bolivia: Can a Social Movement Gov- ern without a Party?" In Woodrow Wilson Center Update on the Americas, No. 2 Democratic Governance and the "New Left": Bolivia: Social Movements, Populism, Presidents and Blank Votes in the Bolivian and Russian Systems 299 and Democracy, ed. Brooke Larson, Raul Madrid, Rene Antonio Mayorga, and Jessica Varat. Miller, Warren. 1955. "Presidential Coattails." Public Opinion Quarterly 19 (4): 353- 68. Nishikawa, Misa, and Erik Herron. 2004. "Mixed Electoral Rules' Impact on Party Systems." Electoral Studies 2 3 (4): 7 5 3-68. Nohlen, Dieter, and Philip Stover, eds. 2010. Elections in Europe: A Data Handbook. Baden-Baden: Nomos. Przeworski, Adam, and Henry Teune. 1970. The Logic of Comparative Social Inquiry. New York: Wiley. Rose, Richard, Neil Munro, and Stephen White. 2001. "Voting in a Floating Party System: The 1999 Duma Election." Europe-Asia Studies 53 (3): 419-43. Sakwa, Richard. 2005. "The 2003-2004 Russian Elections and Prospects for Democracy." Europe-Asia Studies 57 (3): 369-98. Samuels, David J., and Matthew S. Shugart. 2010. Presidents, Parties, and Prime Min- isters: How the Separation of Powers Affects Party Organization and Behavior New York: Cambridge University Press. Seligson, Mitchell A., et al. 2006. "Democracy Audit: Bolivia 2006 Report." http:// www.vanderbilt.edu/lapop/bolivia/2006-audit.pdf. Accessed March 23, 2013. Singer, Matthew M., and Kevin M. Morrison. 2004. "The 2002 Presidential and Parliamentary Elections in Bolivia." Electoral Studies 23 (1): 172-82. Van Cott, Donna Lee. 2008. Radical Democracy in the Andes. New York: Cambridge University Press. White, Stephen. 2005. "The Political Parties." In Developments in Russian Politics 6, ed. Stephen White, Zvi Gitelman, and Richard Sakwa. Durham: Duke Univer- sity Press. CONCLUSIONS Mixed-Member Systems Embedded within Constitutional Systems Chi Huang A concluding chapter may be expected to pull all the previous chapters together. Although inferring from individual chapters generally applicable statements about the consequences of mixed-member electoral systems is not easy, we can still expound upon the common threads running through the entire volume and briefly summarize the major findings of our puzzle- solving efforts. Theoretically, the most important theme that has emerged from this volume is that the effects of electoral systems must be considered within a specific context. Whereas the conventional approach is to consider the effects of legislative electoral systems in isolation, hence called legislature- centric, this volume has repeatedly demonstrated that similar electoral sys- tems embedded within different constitutional arrangements can produce very different sets of incentives. Specifically, researchers should keep both legislative seats and executive offices in view while analyzing the effects of legislative electoral systems. Institutions matter, but the institutional con- text within which those institutions are embedded also matters. Starting Point: A Puzzle Grofman (2001) suggests that political science in general and comparative politics in particular can advance by solving real-world puzzles. This book 300 Conclusions 301 starts with a puzzle for researchers of electoral systems: Why is it that two similar East Asian democracies, Taiwan and Japan, both switched from sim- ilar SNTV to similar new MMM systems, yet displayed divergent speeds and degrees of approaching theoretically expected political consequences (see Huang 2011)? As Huang, Kuo, and Stockton point out in chapter 1, after the electoral reform in 1994, Japan saw a gradual evolution from a multiparty system toward a two-party system over five House of Repre- sentatives election cycles. In contrast, after Taiwan's constitutional amend- ment in 2005, the party system changed almost immediately to a two-party mold in the first postreform legislative election in 2008 and remained so in the second election in 2012. Comparative Methodology The approach we have adopted in this book to solve the puzzle is to look beyond legislative electoral systems per se. Besides examining the similari- ties and differences between each country's electoral rules, we further embed each legislative electoral system within its broad political contexts, especially its constitutional framework of executive power (see also Grofman 1999). At first glance, this approach seems to complicate our task since each country tends to have a unique combination of electoral arrangements, constitutional set-ups, and party systems. Yet this broader perspective forces us to conduct comparative studies both within and between coun- tries. By tracing the same country over time before and after the electoral reform, we can focus more on the pure effects of electoral system change under the identical constitutional framework, although within-case design also tends to lose sight of the inherent interaction between electoral sys- tems and constitutional arrangements. This shortcoming is overcome by drawing comparisons between systems so as to isolate and verify the best candidate, that is, the coordination for pursuing executive offices, to explain their differences in outcomes. The 10 chapters in this volume adopt, explicitly or implicitly, this joint within-system and between-system comparisons approach. Solution to the Puzzle: Executive-Centric Theory Through elaborate country comparisons, all the contributors reach the same conclusion. That is, the electoral system is no doubt important since 302 Mixed-Member Electoral Systems in Constitutional Context it determines how votes cast in an election are translated into seats in the legislature. Yet it does not operate in vacuum. Since parties and elites aim at winning both executive offices and legislative seats, whether the head of government is directly elected and how cabinet posts are allotted also affect how the legislative electoral system shapes politicians' incentives. There- fore, the competition for executive power shapes the party system as much as it does the legislative arena. To fully understand the effects of an electoral system, we must imbed it within the broader political contexts, especially the constitutional framework, of the country in question. The reason for this is because the appeal of and the competition for the executive offices often occupy a higher priority for political parties and elites, and it is the combination of the legislative electoral system and the constitutional set- up that defines the payoffs of capturing executive offices and the degree of cross-district coordination required to capture the executive office. To evaluate the impacts of the mixed-member electoral systems, several factors should be taken into account. Legislature-centric theory indeed spells out clearly the impact of district magnitude, the effective threshold for the party list tier, the ballot structure, assembly size, and dual candidacy, among other factors. Executive-centric theory, on the other hand, further broadens our theoretical landscape by examining "embedded effects" of elec- toral systems, including the payoffs of the executive offices, the degree of coordination required to capture the executive offices, and the sustainabil- ity of the executive offices (e.g., term limits). Other things being equal, the higher the payoffs and the greater degree of cross-district coordina- tion required to win the executive offices, the stronger the incentives for elites to form and stay in the major parties, and therefore the faster the country will move toward a two-party system. As Lin argues in chapter 2, this explains why Japan under the parliamentary system, which allows for some leeway in postelection bargaining, took a much longer time to evolve into a two-party system, while Taiwan under the semipresidential system, in which the winner of the presidential race takes all the cabinet posts and leaves no room for postelection bargaining, moved almost immediately to a two-party system after shifting from SNTV to MMIM. However, the comparative studies looking beyond Japan and Taiwan in part II caution us not to jump to a conclusion that the presidential/ semipresidential system under MAM hastens a two-party system while the parliamentary system does otherwise. As Hicken points out in chapter 8, parliamentary Thailand moved from a multiparty system to a two-party system after adopting an MAIM system in 1997 while the presidential Conclusions 303 Philippines changed in the opposite direction in the post-Marcos era even before implementing an MA/M system. This seemingly contradictory pair serves as a warning against treating the institutional rules mechanically and directs our attention to the underlying factor: the nature of the executive office, including the appeal of capturing the office and the sustainability of the office. Instead of generating a new puzzle, the cases of Thailand and the Philippines actually reinforce the executive-centric theory. The same argument extends to a different type of mixed-member system-mixed-member proportional (MMP). As Tan and Shugart argue in their chapter 9 on parliamentary New Zealand, since MMP was adopted in 1993 the effective number of electoral parties rose between 1993 and 1996 and then has generally declined since the first MMP election in 1996 but a high number of very small "microparties" remains. This stands out in contrast to the mostly similar case of Germany. Taking into account not only the features of the MMP system but also politicians' pursuit of ministerial posts in the cabinet can help illuminate this idea. Batto, Kim, and Matukhno's chapter 10 on Bolivia and Russia further indicates that the so-called contamination effect in the literature of MA/FM systems is not limited to the two ballots for legislators but may well also exist between executive and legislative ballots. In other words, what the voters know of the presidential candidate may also shape their vote choices on the SMD and PR ballots. They find that while voters may vote for the PR list to sup- port the presidential candidate, whether or not these contamination effects extend to the nominal tier depends on the level of institutionalization of the party. This implies that the coattail effects often considered unique to presidential systems combined with single-member congressional districts may also apply in mixed-member systems due to the visibility of the presi- dential candidates, another example of executive-centric theory. Extensions to Other Parts of the Political Systems The number of parties is one of the most heavily researched variables in comparative politics, but other facets of the political system are also crucial in determining the ways in which politics operate in any country. In this volume, contributors have examined faction systems, nominations, post allocation, preelectoral coalitions, and ticketsplitting and found that the embedded institutional context produces interesting variations in each of these areas. 304 Mixed-Member Electoral Systems in Constitutional Context Faction Systems In parliamentary Japan, the change from SNTV to MMN shifted the focus of competition away from intraparty competition and toward interparty competition. As many expected, this did somewhat weaken the LDP's internal factions and strengthen the party leadership. However, Kobayashi and Tsukiyama argue in chapter 3 that the prime minister's need to main- tain parliamentary confidence gives the factions leverage and ensures that they will continue to exercise considerable influence. In contrast, Batto and Huang find in chapter 4 that electoral reform has had a much more mod- est influence on Taiwanese factions. KMT local factions are constructed to win control of local governments, especially city mayors and county mag- istrates, so the change in the national legislative electoral system has not affected them much. DPP factions are more focused on national politics, but the evolution of those factions from legislative-centered to presidential contender-centered is primarily a result of the presidency becoming an achievable goal rather than because of electoral reform. The experience of the KMT and DPP factions also points to the limits of focusing solely on the electoral system in the national legislature, as both have been shaped by a variety of electoral rules, including those for local executives, local assemblies, and party offices. Candidate Selection, Post Allocation, and Preelectoral Coalitions The centrality of executive offices also explains some differences in intra- party competition, nominations, and preelectoral coalitions. Yu, Shoji, and Batto suggest in chapter 5 that because legislators are commonly appointed to executive positions in Japan but not in Taiwan, the incentives to remain in the parliament are much stronger in Japan. The resulting different empha- sis on incumbency implies different sets of problems that parties need to resolve in the nomination stage. In Japan, where the best districts are usu- ally occupied by entrenched incumbents, parties developed the innovative kobo system to allocate nominations for less desirable districts. In Taiwan, incumbents are not protected, and the parties developed a system of poll- ing primaries to determine hotly contested nominations in very desirable districts. Nemoto and Tsai point out in chapter 6 that, in both Taiwan and Japan, parties used cabinet positions to reward party politicians for going out and winning votes in elections. However, while the use of cabinet posi- Conclusions 305 tions as carrots was widespread in Japan, it was much more limited in Tai- wan due to the fact that legislators could not concurrently serve as minis- ters. Likewise, they argue that the presence of institutional features such as the confidence vote and bicameralism made alliances between large and small parties more tenable in Japan. Voters' Behavior The differences in the feasibility of small parties, nomination strategies, and preelectoral coalition that result from differences in the constitu- tional structures in turn define the choice set available to voters and thus shape their patterns of straight- and split-ticket voting. As Wang, Lin, and Hsiao find in chapter 7, in both Taiwan and Japan ticket-splitting takes place mainly within the two major coalition camps rather than across camps. Identifiers with small parties are more likely to be ticket- splitters when they vote sincerely in their PR ballot but support viable district candidates of major parties of the same alliances. This result of voter-level analysis meets the expectation of the conventional legislature- centric theory well. However, it does not mean that the legislature-centric perspective is sufficient to explain ticket-splitting patterns. Quite the contrary, it further confirms the importance of understanding how pre- electoral coalitions are shaped to begin with by both executive posts and legislative seats, a point our executive-centric perspective emphasizes. In essence, political elites structure the alternatives based on institutional incentives-executive as well as legislative-and then voters respond to the available choice set on ballots based the electoral rules. In this respect, executive-centric theory reminds researchers to pay close atten- tion to how both constitutional and electoral systems impose restraints on the way voters can vote. Although this volume focuses on the effects of mixed-member elec- toral systems, especially MA/M systems, we believe that our findings and conclusions have profound implications for the study of electoral systems in general. Electoral systems, important as they are, do not operate in a vacuum. Their origins, evolution, and impacts can be fully understood only when they are embedded within their constitutional frameworks. The literature on electoral rules is indeed impressive and often con- sidered as one of the most advanced areas in political science. Explicitly incorporating executive-centric theory into electoral systems research may be the critical next step leading to future scientific breakthroughs. 306 Mixed-Member Electoral Systems in Constitutional Context Implications for the Quality of Democracy This book mainly focuses on describing and explaining why similar mixed- member electoral systems embedded within different constitutional sys- tems can produce quite different impacts on the party systems and faction systems. Yet the executive-centric theory and the empirical evidence of this volume also have important implications for concern about the quality of democracy, both inside and outside academia. There have been ongoing debates on the trade-off between democratic accountability and representation. There are also advocates of political reforms through electoral engineering. Yet most of the discussions and debates are either on the relative merits of different democratic regimes or on the advantage and disadvantages of various legislative electoral systems. For example, Lijphart (1999) argues that the parliamentary system is supe- rior in terms of the quality of democracy and democratic representation, while Persson, Roland, and Tabellini (1997) find that presidential system is superior in term of fostering accountability. Meanwhile, it is well known that the more "permissive" an electoral system is (such as the PR system), the easier it is for preexisting social cleavages to manifest as political par- ties and thus promote representation. The more "restrictive" electoral sys- tems (such as the SMD system), on the other hand, tend to constrain the number of parties and thus clarify political responsibility (Taagepera and Shugart 1989). Mixed-member systems, despite its hope for getting the "best of both worlds," unavoidably also inherit the advantage and disad- vantages of either according to the way SMD and PR tiers are combined. MMM, which weighs heavily in favor of the SMD tier and against the PR list, tends to strengthen accountability at the cost of representation, while MMP, which allocated seats by the PR ballot, is relatively balanced between the two criteria of the quality of democracy. Although these arguments are enlightening, they tend to miss the key link between the constitutional arrangements and electoral systems. This book bridges the gap by embedding electoral systems within their constitutional systems. The executive-centric theory argues that politi- cal parties and elites seek not only to maximize seats in the parliament but also to seize executive offices. It is the combination of the legislative electoral system and the constitutional set-up that defines the payoffs of capturing executive offices and the degree of cross-district coordina- tion required to win the executive offices. Hence to fully understand the effects of an electoral system, we must imbed it within the broader Conclusions 307 political context, especially the constitutional framework, of the coun- try in question. For example, in presidential and semipresidential systems the president exercises the executive authority. The ultimate goal of most political parties and their leaders is undoubtedly to control both executive and legislative branches, although the appeal of and the competition for the presidency often occupies a higher priority (Curini and Hino 2012). This implies that presidential contests often spill over to legislative elections (Huang and Wang 2014). Thus the conventional wisdom that the separation of power offers voters a better opportunity to hold politicians in different branches accountable for different things deserves further close scrutiny under the executive-centric perspective. Implications for Electoral Reform Electoral systems do not change often, but when they do they are often motivated by high hopes of solving some political problems and enhanc- ing the quality of democracy. But the executive-centric theory of this book implies that electoral reform alone is not a panacea. Whether and to what extent electoral reform achieves the advocated goals depends not only on the specification of new electoral rules per se but also on the context within which such rules are embedded (see also Bowler and Donovan 2013). If a real-world electoral reformer/engineer wants to assess whether their goals are realistically attainable or not, be they reducing money politics, down- sizing factions, or enhancing electoral accountability, she needs a more comprehensive theory to predict how politicians and voters would actually react to such reform. This book brings in a large part of the political pay- off, that is, executive offices, that reformers, just like academics, have not always kept clearly in mind. For example, one of the goals of electoral reform in Japan is to break the importance of factions in order to shift the focus toward interparty com- petition in policy issues. As Kobayashi and Tsukiyama point out in chapter 3, changing from SNTV to MMNIM did somewhat weaken the internal fac- tions of the LDP, but this trend has been slower and far less intense than initially expected. If reformers had taken into account the parliamentary system and the importance of cabinet posts, they would have been less surprised by this result. In fact, the strengthened prime minster even has found room to maneuver within the faction system, preserving the overall 308 Mixed-Member Electoral Systems in Constitutional Context system while enhancing his own power by favoring his own faction and giving posts to smaller rival factions. In New Zealand, on the other hand, the primary aim of the shift from pure SMD to MMP was to increase the quality of representation by encouraging stronger ties between represen- tatives and voters. Shugart and Tan find in chapter 9 that the results are indeed more reflective of the voters' preferences. Yet as mentioned earlier, the effective number of electoral parties has generally declined but a high number of very small "microparties" remain. This has contributed to one of the more unpopular facets of the new system, that smaller parties are often the kingmakers in coalition politics. This fragmentation of the party system can be understood only by taking account of politicians' pursuit of ministerial posts in the cabinet under New Zealand's parliamentary system. Thailand and the Philippines constitute another pair of examples. Elec- toral reformers in both countries desire a stronger nationalized party sys- tem. Yet as Hicken points out in chapter 8, Thailand was successful in mov- ing toward a nationalized politics and a two-party system after adopting the MMM system while the Philippines failed. Limiting the president to one term in the Philippines provides a disincentive for presidential candidates to build strong nationalized parties. Again, the competition for executive offices is important and mediates the effects of electoral systems. In short, the executive-centric perspective advanced in this book mat- ters in the world beyond academia. Political pundits, politicians, reformers, and the general public alike can gain greater insight by keeping both legis- lative and executive payoffs in mind while deliberating the relative merits of different democratic institutions. REFERENC ES Bowler, Shaun, and Todd Donovan. 2013. The Limits of Electoral Reform. Oxford: Oxford University Press. Curini, Luigi, and Airo Hino. 2012. "Missing Links in Party-System Polarization: How Institutions and Voters Matter." Journal of Politics 74 (2): 460-73. Grofman, Bernard. 1999. "Preface: Methodological Steps toward the Study of Embedded Institutions." In Elections in Japan, Korea, and Taiwan under the Single Non-Transferable Vote: The Comparative Study of an Embedded Institution, ed. Ber- nard Grofman, Sung-Chull Lee, Edwin A. Winckler, and Brian Woodall. Ann Arbor: University of Michigan Press. Grofman, Bernard, ed. 2001. Political Science as Puzzle Solving. Ann Arbor: Univer- sity of Michigan Press. Huang, Chi. 2011. "Political Consequences of the MMM Electoral Systems in Tai- wan and Japan." Paper presented at the 2011 Annual Meeting of the American Political Science Association, Seattle, September 1-4. Conclusions 309 Huang, Chi, and T. Y. Wang. 2014. "Presidential Coattails in Taiwan: An Analysis of Voter- and Candidate-Specific Data." Electoral Studies 33 (March): 175-85. Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Performance in Thirty-Six Countries. New Haven: Yale University Press. Persson, Torsten, Gerard Roland, and Guido Tabellini. 1997. "Separation of Pow- ers and Political Accountability." Quarterly Journal of Economics 112 (4): 1163- 1203. Taagepera, Rein, and Matthew Soberg Shugart. 1989. Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven: Yale University Press.  Contributors Nathan F. Batto is an Assistant Research Fellow at the Institute of Politi- cal Science, Academia Sinica and a jointly appointed Assistant Research Fellow at the Election Study Center, National Chengchi University. His research focuses on electoral systems, legislative politics, and Taiwanese politics. His research has appeared in journals such as Electoral Studies, Jap- anese Journal of Political Science, and Legislative Studies Quarterly. Gary W. Cox is William Bennett Munro Professor of Political Science at Stanford University. In addition to numerous articles in the areas of legislative and electoral politics, Cox is author of The Efficient Secret (win- ner of the 1983 Samuel H. Beer dissertation prize and the 2003 George H. Hallett Award), coauthor of Legislative Leviathan (winner of the 1993 Richard E Fenno Prize), author of Making Votes Count (winner of the 1998 Woodrow Wilson Foundation Award, the 1998 Luebbert Prize, and the 2007 George H. Hallett Award), and coauthor of Setting theAgenda (winner of the 2006 Leon D. Epstein Book Award). A former Guggenheim Fellow, Cox was elected to the American Academy of Arts and Sciences in 1996 and the National Academy of Sciences in 2005. Allen Hicken is Associate Professor of Political Science, a Research Asso- ciate Professor at the Center for Political Studies, and Director of the Center for Southeast Asian Studies at the University of Michigan. He is the author of Building Party Systems in Developing Democracies (Cambridge University Press) and the editor of two volumes: Party and Party System 311 312 Contributors Institutionalization in Asia (Cambridge University Press, coedited with Erik Kuhonta), and Politics of Modern SoutheastAsia: Critical Issues in Modern Poli- tics (Routledge). His articles have appeared in the American Journal of Politi- cal Science, Journal of Politics, Comparative Political Studies, Journal of East Asian Studies, Asian Survey, and Electoral Studies. Yi-ching Hsiao is an Associate Professor in the Department of Pub- lic Administration, Tamkang University. His research focuses on voting behavior, congressional politics, and public opinion survey research. His research has appeared in Taiwanese journals such as the Journal of Elec- toral Studies, Taiwanese Political Science Review, Taiwan Democracy Quarterly, Political Science Review, Chinese Political Science Review, and Soochow Journal of Political Science. Chi Huang is a University Chair Professor of Political Science and Research Fellow of the Election Study Center at National Chengchi University. His research interests focus on survey research, quantitative methodology, elec- toral systems, and voting behavior. He has coauthored or coedited six books and published many articles in leading international and Taiwanese journals, including American Political Science Review, American Journal of Political Science, Journal of Politics, Comparative Political Studies, Electoral Studies, Party Politics, Political Research Quarterly, Japanese Journal of Electoral Studies, Issues & Studies, and Taiwanese Political Science Review. He served as President of the Taiwan- ese Political Science Association (2003-5) and has been the Coordinator of the Taiwan's Election and Democratization Study (TEDS) interuniversity survey project since its inception in 2000. Hsin-ta Huang is an Assistant Professor in the Department of Political Sci- ence at Tunghai University. He received his PhD from National Chengchi University. His research interests include political methodology, quantitative analysis, local factions, party systems, electoral systems, and voting behavior. Henry A. Kim is an unaffiliated researcher with a PhD in political science from the University of California, San Diego. His research includes elec- tions, parties, legislative politics, and the cognitive psychology of decisions under uncertainty. His publications have appeared in State Politics and Policy Quarterly, Japanese Journal of Political Science, and American Journal of Politi- cal Science. Yoshiaki Kobayashi is a Professor of Political Science at Keio University and a Vice President of Science Council of Japan, Prime Minister's Office of the Contributors 313 Japanese Government. He serves on the editorial board of several interna- tional journals including International Political Science Review, Pacific Focus, and JPSS. His research interests focuses on democracy in civil society, particularly political attitudes, electoral behavior, and electoral systems. He has authored 12 books in English, Japanese, Korean, and Chinese, and coauthored or coed- ited 51 books and 300 articles in leading International and Japanese journals. He has served as President of the Japanese Political Science Association, the Japanese Association of Electoral Studies, and the Japan Public Choice Society. Ming-Feng Kuo is a Research Fellow at the School of Public Admin- istration, Sichuan University, China. His research interests include Japa- nese politics, comparative politics, public management, policy analysis, and quantitative methodology. His research has appeared in journals such as Administration & Society, Taiwanese Journal of Political Science, Journal of Social Sciences and Philosophy, Journal of Electoral Studies, Chinese Political Sci- ence Review, Journal of Civil Service, and Wenti Yu Yanjiu. Chang-chih Lin is an Adjunct Assistant Professor of Political Science at National Chengchi University. His research interests focus on compara- tive politics, electoral systems and party politics, voting behavior, and sur- vey research. His research has appeared in journals such as the Journal of Electoral Studies, Taiwanese Political Science Review, Issues & Studies, Wenti Yu Yanjiu, and Chinese Political Science Review. Jih-wen Lin is Research Fellow in the Institute of Political Science at Academia Sinica and Professor of Political Science at National Chengchi University. He received his PhD in political science from the University of California at Los Angeles (1996). His research interests cover game theory and comparative institutional design, and he has had articles published in Party Politics, Journal of Democracy, Electoral Studies, China Quarterly, Journal of EastAsian Studies, and Issues & Studies, among others. Natalia Matukhno is Research Fellow at the Centre for the Study of Pub- lic Policy at the University of Strathclyde. She received her PhD in politi- cal science from the University of Arizona in 2013. Her research interests include comparative political institutions and comparative political behav- ior, particularly for democratizing countries. Her dissertation focuses on the varieties of corruption. Kuniaki Nemoto is Assistant Professor at Waseda University, Japan. He previously taught at the Department of Political Science, University of 314 Contributors British Columbia, Canada, and was Korea Foundation Postdoctoral Fellow at the Korea Institute, Harvard University. He earned his PhD in politi- cal science and international affairs from the University of California, San Diego. He is interested in elections, party politics, and legislatures, with a regional focus on the Asia-Pacific. His articles have appeared in the Brit- ish Journal of Political Science, Comparative Political Studies, Party Politics, and Electoral Studies. Hans Stockton is Director of the Center for International Studies at the University of St. Thomas in Houston, Texas, Chair of the Department of International Studies, and holder of the Sarofim-Cullen Trust Endowed Chair in International Studies. His areas of academic specialization are democratization, elections, and security in contemporary Asia Pacific. He has published in journals of political science, economics, public administration, and Asian Studies; authored multiple chapters in edited volumes on military affairs in the Asia Pacific and Taiwan's political development; and authored two books. He currently serves as President of the Southwest Conference on Asia Studies, Vice President of the American Association of Chinese Studies, and Coordinator of the Conference Group on Taiwan Studies. Kaori Shoji is a Professor in the Department of Political Studies at Gakushuin University in Japan. Her research focuses on party organiza- tions, electoral systems, policy innovation, and American politics. Matthew S. Shugart is Professor of Political Science at the University of California, Davis. He is the coeditor of Mixed Member Electoral Systems: The Best of Both Worlds? (Oxford, 2001) and coauthor of Presidents, Par- ties, and Prime Ministers: How Separation of Powers Affects Party Organiza- tion and Behavior (Cambridge, 2010). His articles on electoral systems have appeared in the American Journal of Political Science, Electoral Stud- ies, and other journals. He acknowledges the Research School of Asia and the Pacific, Australian National University, and the U.S. National Science Foundation (grant SES-0751662) for supporting field research in New Zealand in November 2011. Alexander C. Tan is Professor of Political Science at the University of Canterbury in Christchurch, New Zealand. He has published works in the areas of East Asian politics, comparative political economy, and political parties, and his articles have appeared in journals such as Journal of Politics, Contributors 315 Political Research Quarterly, Comparative Political Studies, Comparative Politics, European Journal of Political Research, Electoral Studies, and Party Politics. Chia-hung Tsai is Director and Research Fellow at the Election Study Center, National Chengchi University. He served as a visiting faculty member in the Department of Political Science, Massachusetts Institute of Technology, in 2012-13. His research focuses on public opinion, voting behavior, and methodology. He has published research in journals such as Party Politics, International Political Science Review, and Issues & Studies. Hiroki Tsukiyama is a research associate at Keio University. His research interests include Japanese politics and legislative politics. He received his PhD in Political Science from Keio University. His dissertation explores the legislative process of Japanese local legislatures. T. Y. Wang is Professor of Political Science at Illinois State University. He currently serves as the coeditor of the Journal of Asian and African Stud- ies and was the Coordinator of the Conference Group of Taiwan Studies (CGOTS) of the American Political Science Association. Professor Wang's current research focuses on Taiwanese national identity, cross-Strait rela- tions, Chinese politics, electoral studies, U.S. policy toward China and Tai- wan, and research methodology. He has authored, coauthored, or edited 6 books/special issues and published more than 30 articles/book chapters in such scholarly journals as the American Political Science Review, Asian Sur- vey, International Studies Quarterly, Issues & Studies, Journal of Peace Research, Political Research Quarterly, and Social Science Quarterly. Eric Chen-hua Yu is an Associate Research Fellow of the Election Study Center and jointly appointed as an Associate Professor of Political Science at National Chengchi University in Taiwan. His research interests include public opinion, electoral politics, quantitative methods, and American poli- tics. Yu recently published academic articles on Taiwan's and U.S. domestic politics in Taiwan Political Science Review, Journal of Electoral Studies, Review of Social Sciences, Japanese Journal of Electoral Studies, and Japanese Journal of Political Science. Yu received a MS (2000) in Public Policy Analysis from the University of Rochester, and a PhD in political science (2006) from Columbia University.  Index Acci6n Democratica y Nacionalista (Bolivia), 287, 292, 294-95 ACT (New Zealand), 249, 250, 257-64, 274-75 Alliance (New Zealand), 248, 250, 253, 257, 259, 273 average effective number of electoral par- ties in each district (N11), 236, 240-41 best loser provision, 27, 33, 42-44, 167, 177, 182, 189, 270 bicameralism, 54-55, 57, 62, 66-67, 183- 84, 233, 305 bifactionalism, 14,103-9,113,116,120, 122,128 bipolar competition (bipartisan, bifac- tional), 5, 12 blank votes (against all votes), 278-80, 282-87, 289, 291-97 block vote, 69, 230-31, 236, 243 Bundestag, 254, 274 cabinet appointment, 172-75, 189 cabinet portfolios (ministerial posts, executive positions), 2, 12, 16, 20, 53-56, 58-59, 61-63, 66-69, 99, 244, 257, 260-61 campaigning, 61, 97, 155, 172, 251, 279 candidate coordination, 55, 166-67, 182, 188-90 candidate selection method (nomina- tions), 2, 7-8, 14-15, 17, 30, 46, 73, 75-76, 81-82, 84, 95, 97, 109, 117-18, 120,124-26,132,135-46,148-56, 158-62, 166, 175, 177, 181-82, 188- 89, 191, 196-98,201-2,207-8,210, 212, 218, 220, 288, 303-4 Chen Shui-bian, 7, 9,14, 31, 59, 60, 119-20,173,187 Christian Democratic Union (Germa- ny), 254, 273 Christian Social Union (Germany), 254, 273 Clean Government Party (Komeito) (Japan), 33, 45, 46, 183-86, 188, 190,199-200, 202-5,207-10,212-14,217,220 closed list PR, 29, 33, 69, 74, 287 closed primaries, 149, 157, 159-60 Communist Party of the Russian Fed- eration, 289, 291-92, 294-95 constitutional reform, 191, 236 contamination effects, 42-43, 67, 188, 248, 265-66, 269, 278-79, 295, 303 coordination (role of political parties), 2-3,10-11,19,27,29,46,54-55,61, 76, 127, 151, 162, 166-67, 172, 182- 83,188-90,196,201,219,229, 317 318 Index coordination (continued) 232-36, 238, 240-41, 243, 252, 301- 2,306 county assembly, 9, 13, 103, 109-12, 131 county magistrate (city mayor, local executive), 103, 107-9, 112-13, 129 cross-district coordination, 229, 232, 236,240-41,243,302,306 democratic accountability, 306 Democratic Party of Japan, 33, 37, 65, 66, 80, 135, 138-39, 152-63, 165, 168, 176-92, 199-220 Democratic Progressive Party (Taiwan), 7-8, 11-15, 17, 20, 28, 31, 38, 60, 102- 31,135-50,161-62,171-73,178-82, 186-91, 199-217, 304 Diet, 12, 28, 30-31, 33, 36, 40,45, 53- 55,57,62-63,74-80,89,91,94-95, 97-98, 151, 153, 168-70, 176, 199- 200, 238 Diet committees, 77, 168, 170 district magnitude, 7, 9, 28, 32, 34, 136, 177, 190, 232-33, 236, 274, 302 dual candidacy, 15, 27, 33, 42-44, 161, 167-68, 176-77, 189, 249, 251, 269- 73, 302 Duverger s Law / Duvergerian incen- tives, 36, 38, 182, 186, 197, 205-7, 218-19, 248, 252, 265, 268 effective number of electoral parties (Nv), 36-37, 43, 182-83, 233, 240, 243, 247, 266-69, 275, 303, 308 effective number of electoral parties in PR tier (NPR), 36-38 effective number of electoral parties in SMD tier (NSMD), 36-38, 45, 236, 239-41, 244 effective number of electoral parties in SNTV tier (NSNTV), 36-3 7 effective number of parliamentary par- ties (Ns),6, 36-39, 41-43, 253-55, 273-74 effective number of parties, 3, 6, 32, 36- 37, 53, 58, 67, 199, 232-33, 236-37, 240-41, 245, 247-49, 253-54, 256, 258,266,272 effective number of presidential candi- dates (Npre), 242, 245 Electoral Referendum Act, 270, 306 electoral reform, 1, 3-10,12,15-17, 26-28, 30-31,35-36,38, 41,44,52- 53, 57-58, 62, 66, 68-69, 74, 76-81, 83-85, 89-91, 93-95, 97-100, 102-3, 118, 123,127, 135-36, 138, 143, 151-52,156,160-61,166,169,179, 198,218,232,287-88,298,301,304, 307-8 endorsement. See candidate selection methods executive-centric theory, 2, 27, 52, 301- 3,305-7 executive-legislative relations, 2, 27, 52, 301-3,305-7 faction proportionality system, 85, 89, 91,93-95 faction systems, 1, 3, 5, 12, 16-17, 103, 107, 128, 303-4, 306 factionalism, 6, 9,13,20,30-31, 68, 73-75, 78, 103, 106, 113, 116, 127-28, 166,173,182,233-35,238 female candidates, 160, 176, 181 Foreshore and Seabed Act (New Zea- land), 258 Formosa faction (Taiwan), 13-14, 118- 21, 125-28 fragmentation, 18, 52, 166-67, 185, 237, 239-41, 243-45, 249, 266, 268, 308 Free Democratic Party (Germany), 254, 273 fused vote, 286 Gallagher s Least Squares Index, 250-51 governing coalition, 199-200, 209, 211, 213-14 Green Party (Germany), 254 Green Party (New Zealand), 248, 250, 252-54, 257-58, 262-63, 270-71, 273, 275 guanxi, 20, 105-6 Hatoyama, Yukio, 64-66 head voters (rentou dangyuan), 140-41, 149 Index 319 Hosakawa, Morihiro, 31 House of Councillors, 31, 53-57, 62-68, 156, 185 House of Representatives, 31, 34-38, 53-55,62-63,66-69,137,152-60, 163,194,220,230-31,235,238, 301 inchoate parties, 19, 282-85, 294 information (for voters), 37, 61, 82, 87, 91, 97-98, 155,213,214-16,219, 278, 280-85 institutionalized parties, 279, 281, 283, 285, 291, 293, 296 interparty coalitions, 7, 9, 14-15, 17, 20, 116, 118, 166-67, 182, 184, 186, 188- 91,196,198,204,208,211,218,257, 261, 263, 303-5 intraparty competition, 30-31, 76, 95, 135-36, 143-44,153,304 Japan Restoration Party, 33, 64, 152, 159, 186, 191-92 Japan Socialist Party, 10-11, 19, 28, 64, 138, 152-53 Japanese Communist Party, 33, 200, 202-3, 205, 207-10, 214, 217, 220 joint PR list, 183, 187 Justice Alliance (Taiwan), 14, 119-22, 124-27, 131 kenren (prefecture party branch), 153-57, 159, 163 kobo system (public recruitment), 17, 135-36,138-39,150,152-63,175, 304 Koizumi, Junichiro, 64, 77-79, 83, 85, 87, 90-92, 94-95, 98-99. 176, 190 Komeito. See Clean Government Party Kuomintang. See Nationalist Party Labour Party (New Zealand), 257-59, 262, 275 Left Party (Germany), 248, 254 legacy politicians, 249, 2 58-59, 261, 272 Legislative Yuan, 10, 31-32, 40-41, 54, 58-60, 68, 70, 109, 112, 121, 135, 141- 43, 145, 161, 172-73, 179,219 legislative-centric theory, 52, 232, 302, 305 Liberal Democratic Party (Japan), 6-8, 13-15, 17, 27-46, 64-65, 73-100, 102, 116,125-39,151-61,165-70,176-91, 199-220,290,307 Liberal Democratic Party of Russia, 290-92, 294-95 Liberal Party (Philippines), 239, 241 list ranking, 176-77, 189 list tier, 4-5, 18-19, 27, 32-35, 42-45, 118,161,167,173,175-77,179, 182-83, 185-86, 189, 231-32, 236, 239,249,265-66,269,272,278-80, 282-86, 291-94, 296-97, 302 list tier threshold, 27, 34-35, 43, 46, 207, 302 Ma Ying-jeou, 9, 59-60, 70, 124, 172-74, 187,201 majoritarian-leaning electoral system, 52-53, 63, 67 Mana Party (New Zealand), 250, 258-59 Maori Party (New Zealand), 250, 255, 257-60, 262, 275 Maori special seats, 258-59, 274 Marcos, Ferdinand, 231, 238, 241-42, 245 mechanical effect. See Duverger s Law microparties, 248, 258-61, 263-65, 270, 303, 308 minimum winning coalition, 91 ministers / vice ministers, 97, 168, 184 MMM characteristics, Japan and Taiwan compared, 6, 27, 30-3 5, 52 most similar systems design, 16, 38, 53, 301 Motherland (Russia), 290-92, 294-95 Movimiento al Socialismo (Bolivia), 287-89, 292, 294-95 Movimiento de Izquierda Revoluciona- ria (Bolivia), 287, 289, 292, 294 Movimiento Indigena Pachakuti (Bolivia), 288-89, 292, 294-95 Movimiento Nacionalista Revoluciona- rio (Bolivia), 287, 292, 294-95 multinomial logit model, 211, 213-17 3 20 Index Nacionalista Party (Philippines), 239, 241 National Assembly (Taiwan), 32, 109, 161 National Party (New Zealand), 251, 258, 264, 271, 273, 275 Nationalist Party (Kuomintang) (Tai- wan), 8-11, 13-14, 17, 19-20, 27-28, 31-32, 59-61, 70, 102-17, 128-30, 135, 139-50, 161-62, 171-91, 199- 219,304 New Party (Taiwan), 20, 31, 34, 200- 201, 203, 205-8, 210, 212, 215-17, 219 New Tide faction (Taiwan), 13-14, 108, 114,118-27,131-32,190 New Zealand First Party, 249, 253, 257- 58,260,262-64,271,275 1955 system (Japan), 27-28, 45, 75 nominal tier, 5-6, 14, 18-19, 32-33, 35, 38,45, 161, 169-70, 175-76, 181-82, 187-89,249,266,273,275,278-80, 282-86,291,303 nominations. See candidate selection methods nonopportunity voters, 202, 205, 207-8, 210-11, 220 nonpartisan ministers, 57, 59-60, 63, 68-70 Non-Partisan Solidarity Union (Tai- wan), 200-203, 207-8 Nueva Fuerza Republicana (Bolivia), 288-89,292,294-95 one-way ANOVA, 59-61, 63, 65, 68 opposition coalition, 185, 199, 209, 213- 14, 220 overhang, 251, 273 Pan-Blue alliance, 31, 38, 110, 186, 188, 201, 205, 209, 211-12, 215-17 Pan-Green alliance, 172, 187-88, 201, 206-7, 209, 212, 215-17 parliamentary system, 13, 16, 38, 53, 62, 67-69,74,95,137,160,167,230,235, 237, 243, 248, 253, 256, 280-81, 295- 96,302,306-8 party factions, 7, 140, 166, 236, 244, 248 party headquarters, 74-78, 81-83, 86-87, 95, 98, 139, 141-42,144-47,152-56, 158, 163, 171 party identification, 196-97, 199, 204-9, 211,218 party leadership, 75, 128, 130, 166, 173, 304 Party of Democratic Socialism (Ger- many), 254, 274 party posts, 15, 17, 73, 75-77, 97, 168 party reputation (party image), 165-66, 169, 236, 283, 285, 296 party system, 5-6, 12, 18, 26-29, 35, 37-38, 41-44, 52-53, 58, 63, 66, 69, 198-99,229-30,232,236-43,245, 248, 263, 265-66, 278, 285, 287, 289, 295,301-2,308 patronage, 73, 104-6, 116-17, 120, 129, 287-89 patron-client relationships. See patron- age People First Party (Taiwan), 11, 20, 34, 38,61,110,112-14,187-88,190, 200-3,205-8,210,212,215-17,219- 20, 297 Peoples New Party (Japan), 66, 152, 184-85, 188, 190, 200, 220 personnel management, 75, 78-79, 90, 95 pivot, 54-55, 63, 261, 275 plurality rule, 1, 7-8, 25, 32, 116, 130, 197, 230-31 Political Funds Control Act reform, 76 polling primary, 17, 135-36, 138-50, 160, 162, 304 portfolio allocation. See cabinet port- folios PR constituency, 31, 33-34, 43-44 preelectoral coalition, 17, 116, 165-91, 303-5 presidential coattails, 167, 186-87, 281, 303 presidential system, 16, 38, 58, 67-68, 137, 160, 231, 238, 278, 280, 283-86, 295-96,302-3,306-7 proportional representation (PR), 1-2, 5,26,29,31,33-36,42-44,46,55,61, 69, 74, 76, 97, 156, 161-62, 168, 175- 77, 183, 185, 187, 191, 195, 197-208, Index 321 210, 212, 218-20, 231-32, 248-49, 265-66,268-70,275,287, 303, 305 psychological effect. See Duverger s Law quality of democracy, 306-7 quasi-experimental design, 26, 35, 38 recruitment, 17, 74, 78-84, 88, 97-98, 135, 151-52, 158, 175, 181,295 representation, 1, 25, 34-35, 69, 74, 97, 195, 198, 231-32, 247, 249-54, 258- 59,261,263-64,266,269-70,272, 274, 306, 308 research design, 35, 285 semipresidential system, 16, 38, 58, 67- 69, 137, 160, 167, 171, 189, 302, 307 Senate, 230-38, 244, 286 seniority, 13, 77, 137-38, 160, 168-69, 264 seshu (hereditary succession), 151, 156- 58, 160 Shinawatra,Thaksin, 231, 243 single-member district (SMD), 1-2, 4, 27, 32-33, 35-38, 42, 44-46, 56, 67, 74, 79, 123, 136, 156, 159, 161, 166- 70, 177-78, 181-83, 185-89, 191-92, 195,197-208,210-20,253,266,268, 274, 303, 306, 308 single-member plurality (SMP), 5, 43, 66-67,183,200,209,248-49,252-54, 259-60,265-73 single nontransferable vote (SNTV), 3-9, 12-15, 17, 19, 26-32, 36-41, 44-46, 55-57, 59-61, 63, 65-67, 69, 73-85,89-95,97,99,102-3,109-10, 112, 114,116-18,125-29,136-37, 143, 151, 160, 165-66, 173-75, 182, 189,191,243,301-2,304,307 Social Democratic Party (Germany), 248,254 split-ticket voting, 17-18, 194, 196, 198, 203,208-9,211-16,220,305 straight-ticket voting, 197, 202, 205, 208-9,211-12,213-16,218,220,305 strategic voting, 36, 42, 138, 207, 219, 282 Taagepera's Seat Product, 255-56, 274 Taichung County Black faction (Taiwan), 103, 107-8, 110-15, 130 Taichung County Red faction (Taiwan), 103,107-8,110-15,128,130-32 Taiwan Solidarity Union, 12, 20, 34, 44, 61, 110, 112-14,174,186-88,190, 200-3,206-10,212,215-17 term limits, 240, 245, 302 Tomorrow Party of Japan, 33, 186, 192 two-party system, 5-6, 18, 26-2 7, 29-3 0, 35,37-38,42,44,52,198-99,202, 230, 237, 266, 301-2, 308 unified and divided governments, 57, 59-60, 219-22 Union of Right Forces (Russia), 290-91, 294-95 United Future (New Zealand), 250, 257- 58,260-63,275 United Russia, 290-92, 294, 295 vote of no confidence, 54,58-59, 94-95, 281 Welfare State Alliance (Taiwan), 14, 114, 119-22, 124, 126, 127 Yabloko (Russia), 290-92, 294-95 Your Party (Japan), 152, 186 zombie, 15, 20, 33, 43, 168, 170, 177, 189,271