Inverse Relationship between Party and Party System Institutionalization: The Transformation of Postwar Japanese Party Politics 1

Inverse Relationship between Party and Party System Institutionalization: The Transformation of Postwar Japanese Party Politics 1 Takayoshi UEKAMI Okayama University Hidenori TSUTSUMI Kagawa University 2016/04/24-28 Correspondence concerning this paper should be addressed to Takayoshi UEKAMI, Okayama University. Contact: uekamit@okayama-u.ac.jp Abstract The aims of this article are to clarify the relationship between institutionalization of political parties and institutionalization of the party system as a whole, and to explore the mechanisms of change in this relationship at a micro level. The subject of institutionalization of parties and the party system has been addressed to date in the course of studies of political parties in the field of comparative political science. This article identifies four patterns in institutionalization and de-institutionalization of parties and the party system. It is theoretically conceivable not only that institutionalization and de-institutionalization may proceed concurrently, but also that, as a party system is institutionalized, political parties themselves may become de-institutionalized; or, conversely, that de-institutionalization of the system is accompanied by institutionalization of parties. A review of postwar Japanese party politics reveals that as the one-party dominant system became institutionalized, the dominant party was de-institutionalized from the inside by factionalism and clientelism among members. The one-party dominant system was replaced by a more competitive party system from the 1990s onwards, the factionalism that formerly held sway within the dominant party was suppressed, and the dominant party became united by conservative ideology. Nevertheless, the institutionalization and de-institutionalization of parties and the party system are macro-level concepts, and lack the capacity to account for micro-level mechanisms of change. This article is therefore based on the hypothesis that the pattern and degree of competition among political parties 1 This paper was prepared for the ECPR workshop on Institutionalization and De-Institutionalization of Political Organizations, held on April 24-28, 2016 in Pisa, Italy. The original paper was delivered to the conference on Institutionalization and De-Institutionalization of Political Organizations held on November 20-21, 2015 at National Sun Yat-Sen University, Kaohsiung, Taiwan. Authors thank to helpful comments from Alex Tan, Lars Svåsand, Rachel Gibson, Frank Liu, Wan-Ying Yang, Ching-Hsin Yu, Yen-Pin Su, Chiung-chu Lin, and Boyu Chen. This work was supported by JSPS KAKENHI Grant Numbers 26380173. 1

exerted influence on the power of the dominant factional coalitions and factions within Japan s Liberal Democratic Party (LDP), leading to change in the party s organization and policies. Analysis of data gathered from members of the House of Representatives reveals that a reshuffle of partners in the coalition government at the end of the 1990s led to the dominant factional alliance within the LDP being replaced, resulting in a shift in LDP policy from moderate to conservative; and that increasing inter-party competition prompted a concentration of power in the party executive that engendered a decline in factionalism. 2

1. Introduction The most prominent characteristic of Japanese party politics in the past was the single-party dominance of the Liberal Democratic Party (LDP). Formed in 1955, the LDP was in power continuously until 1993. It was infamous for factionalism and clientelism, although these characteristics largely disappeared in the 1990s, when the party system became more competitive. Changes of government occurred in 1993, 2009 and 2012. Power relations in the LDP were consolidated, leading to a decline in factionalism and clientelism among members of the Diet, Japan s national parliament. The present LDP, which returned to government following the House of Representatives (HR; Lower House) election in 2012, has been brought together by Prime Minister Abe Shinzō under a policy of conservatism. How should we understand these changes in Japan s formerly dominant political party? In order to answer this question, this paper examines the mechanisms of change on a micro level, based on an analysis of the institutionalization of the party and the party system as a whole. This institutionalization has been discussed previously in the field of comparative political science. This paper applies the concept of institutionalization to gain an understanding not only of emerging democracy in developing countries and new political parties in developed countries, but also of the changes undergone by established political parties in developed countries. First of all, we classify patterns of institutionalization or de-institutionalization of the political party and party system into four types. Theoretically, institutionalization or de-institutionalization of both the party and the party system may occur simultaneously (Type I, IV); the institutionalization of the party system may lead to the de-institutionalization of the party itself (Type III); or the de-institutionalization of the party system may lead to the institutionalization of the party (Type II). Reviewing party politics in Postwar Japan, we show that as the system of single-party dominance became institutionalized, the dominant party became subject to the de-institutionalizing internal forces of factionalism and clientelism (Type III). After the 1990s, as a more competitive party system replaced single-party dominance, factionalism within the former dominant party was suppressed, and an effort was made to bring the party together under a conservative ideology (Type II). The concept of institutionalization or de-institutionalization of the party and party system works on a macro level, however it fails to explain the mechanisms of change on a micro level. We propose the theory that changes in the pattern and degree of inter-party competition influences the power of the dominant factional coalition and factions within the party, thus leading to changes in the party s structure and policies. Using data on Members of the HR, we show how a reshuffle of coalition partners caused changes in the dominant factional alliances within the LDP, and a shift from moderate to conservative policies. In addition, we show how an increase in the degree of 3

competition between political parties caused an increase in the power of the LDP executive, and brought about a decrease in the number of HR Members belonging to a faction. Below, we first discuss the concept of institutionalization or de-institutionalization on a macro level, before proceeding to an examination of the mechanisms of change on a micro level. Finally, we present our conclusions and their implications. 2. The Institutionalization and De-Institutionalization of Party Politics in Postwar Japan In this section we evaluate the institutionalization both of the political party, and of the party system, the changes they have undergone and the relationship between them, based on a concise depiction of party politics in Japan after the end of the Second World War. This principally entails a macro-level analysis, from an historical perspective. According to Harmel and Svåsand (2015), a political party is institutionalized when its actions have become routine, other actors subjectively recognize its lasting power, and said party exhibits objective durability. Furthermore, this institutionalization can advance and retreat in multiple dimensions. Randall and Svåsand (2002) suggests that institutionalization has two dimensions: the dimension of internal and external, and the dimension of structural and attitudinal. Internal structural institutionalization is conceptualized as systemness ; internal attitudinal institutionalization is value infusion ; external structural institutionalization is decisional autonomy ; external attitudinal institutionalization is reification. 2 Party institutionalization will be hindered when these formal institutions compete with informal ones (Harmel and Svåsand op.cit., p.8). Randall and Svåsand s work distinguishes the institutionalization of the political party, and that of the party system, focusing on the former. They suggest however that the two dimensional approach above may also be applied to the institutionalization of the party system, and provide important insights into the relationship between the institutionalization of the political party, and that of the party system (pp.6-9). Figure 1 shows the four combinations of institutionalization of the political party and party system. 2 Basedau and Stroh (2008) measures the institutionalization of African political parties based on the four elements of organization, coherence, autonomy, and roots in society. Arter and Kestilä-Kekkonen (2014) also attempts further operationalization of Randall and Svåsand s concept. To an analysis of the True Finn Party (PS), they apply a number of criteria including: stable support from voters at elections, organizational adequate number of party members and candidates, monopolization of power by the party leader, and adequate order in the parliamentary party. Lindberg (2007), treating institutionalization of the party system as stabilization of interaction among different political parties, undertakes a categorization of party systems in various African nations with a focus on fluctuation the number of parties and seats held. 4

Figure 1. Relationship between party and party system institutionalization Party system institutionalization de-institutionalization institutionalization I Mutually reinforcing II Party de-institutionalization III IV Mutually destructive Type I is where the party reinforces the party system, and vice versa. Conversely, Type IV is where the collapse of the party and party system progress simultaneously. The problem lies in Type II and Type III, where the institutionalization or de-institutionalization does not occur in tandem. Randall and Svåsand suggest that situations where power is unevenly distributed between parties, or where a party is linked to an exclusionary subculture, may cause the institutionalization of the party to have an adverse effect on the institutionalization of the party system. An inverse relationship of cause and effect exists in Type II, where the de-institutionalization of the party system drives the institutionalization of the party. Let s assume that a decline in party support from voters and/or electoral reform can transform the party system 3. In an increasingly uncertain environment, an existing party may choose internal institutionalization in order to survive. Room may also open up for new parties to be formed and institutionalized 4. As the de-institutionalization of the party system progresses, however, the structures and behaviors that surround the party become more fluid, meaning that institutionalization is limited to the party s internal organization. In Type III, it is postulated that within a highly institutionalized party system, the party may become caught up in factional struggles. In this case, de-institutionalization is limited to the intra-party sphere, as the institutionalization of the party system causes the party s external environment to stabilize. We conclude that the single-party dominance enjoyed by the LDP was the combination of an institutionalized party system and a dominant party exposed to internal de-institutionalizing 3 According to Cox (1997), the electoral system and/or social cleavages affect the coordination of elections among candidates, thus determining the number of political parties. 4 Biezen (2003) points out the differences between how existing and new parties form and develop. According to Lago and Martínez (2011), three factors that contribute to the creation of new parties are a short period of time since the first election, a high degree of electoral volatility, and a low electoral threshold. 5

pressures. The political reforms of the 1990s made Japan s party system more competitive, and coalition governments were formed. This de-institutionalization of the party system brought about the weakening of factions and increasing institutionalization within the dominant party. In other words, postwar Japan may be understood as an example of the transition from Type III to Type II. Party Politics in Postwar Japan It was not until 1955 that the fundamental composition of party politics in postwar Japan became established. Numerous political parties appeared after the end of the War, but political camp was gradually consolidated on the conservative side by the LDP, and on the progressive side by the Socialist Party of Japan. The 1960s saw an increase in the number of opposition parties, with the Democratic Socialist Party (a breakaway from the Socialist Party), the Komeito (a party founded by a newly formed Buddhist group), and the Japanese Communist Party (legalized after the War) each gaining seats, especially in urban areas. These opposition parties were policy-seeking: their purpose was the realization of their ideology, and they never threatened the continuation of the LDP government. The LDP, in contrast, was vote-seeking: while advocating a policy of conservatism, it actually placed most importance on the result of popular elections in order to remain in government. The LDP was in government until 1993, with a majority of seats in the House of Representatives. The political opposition was fragmented, with the Socialist Party only ever holding about half as many seats as the LDP at most. The chances of a change in government were slim. In terms of comparative politics, Japan was classified as a one-party dominant political system (Sartori 1976; Pempel ed. 1990). It was not until the 1990s that this pattern of party politics, established in 1955, began to undergo a transformation. According to Randall and Svåsand, party systems become institutionalized when the relationship between parties is stable (internal and structural aspect) and there exists mutual recognition among the parties (internal and attitudinal aspect), as well as when the relationship between the state and the party system as a whole is stable (external and structural aspect) and the electorate supports the party system (external and attitudinal aspect). During this period in Japanese party politics the stability of the relationship between parties, and between the political system, the state and the public electorate, fostered a high degree of institutionalization. Political Reform The electoral system formerly used in the HR was classed as a single non-transferable vote (SNTV) system for a fixed multiple number of seats. This system is known to encourage the casting of personal votes (Carey and Shugart 1995). In addition, local governments in Japan were dependent on allocation of financial resources from the central government. Scheiner (2006) explains that the combination of clientelism and centralization of financial power underpinned the support for LDP 6

governments. In the 1990s, public criticism of this political corruption bolstered efforts for political reform. Immediately prior to the 1993 HR election, dissenting members from the LDP formed two new political parties: the Japan Renewal Party and the New Party Sakigake. In the HR election that followed, the LDP was unable to gain a majority, and the eight political parties excluding the LDP and the Communist Party formed a coalition government. In 1994, the new non-ldp coalition government succeeded in changing the HR electoral system from a SNTV system to an electoral system comprising single-member districts (SMD) and proportionally represented (PR) multiple-seat constituencies, and establishing a system of public subsidies for political parties. Subsequently however, schisms formed in the new coalition due to the formation of new political party, and the coalition soon collapsed. The new electoral system consisted of 300 seats from SMD and 200 seats distributed according to PR by the d'hondt method, thus placing greater weight on SMD. 5 It was necessary for the fragmented opposition parties to work together in order to challenge LDP candidates in SMD. It was in this context that political parties participating in the non-ldp coalition excluded the Socialist Party and New Party Sakigake, which had come into conflict over the operation of government, and formed the New Frontier Party (NFP). The NFP differed considerably from previous opposition parties in that it was a vote-seeking party that viewed winning government as a primary objective. 6 Meanwhile, the LDP entered into negotiations with the two parties excluded from the NFP, the Socialist Party and Sakigake, and succeeded in forming a coalition government. De-Institutionalization of the Party System The LDP regained power at the 1996 HR election, and reclaimed government in its own right, but lost its majority in the House of Councilors (HC; Upper House) at the 1998 election. In the next year, the LDP formed a new coalition government with the Komeito and the Liberal Party. The Liberal Party soon broke away, but the coalition between the LDP and the Komeito lasted until the government s defeat at the 2009 HR election. In contrast, the NFP disbanded after failing to win power at the 1996 HR election. Many members of the former NFP, excepting those originally from the Komeito, joined the Democratic Party of Japan (DPJ), which had been formed by members from the Socialist Party and Sakigake. Before the 2003 HR election, the Liberal Party, which had broken from the LDP coalition, merged with the DPJ, further increasing the DPJ s power base. The main aim of the DPJ was to win government, and in this sense it posed a real threat to the LDP (Uekami 5 In the proportional representation system the electorate is divided into 11 regional blocks, with voters casting a vote for a political party in each block. The electoral system in place today comprises 295 single-seat constituencies and 180 proportional representation constituencies. 6 Strom (1990) suggests that in a competitive two-party system, parties tend to embrace vote-seeking as their objective (p.592). 7

and Tsutsumi eds. 2011). The LDP sought to use traditional clientelism to mobilize supporters (Calder 1988). As budget deficits accumulated, however, the NFP and DPJ called for radical administrative reform. In this context, Koizumi Jun ichirō, who became President of the LDP and Prime Minister of Japan in 2001, gained the support of the electorate for his policy of economic reform based on neo-liberalism (Otake 2003). Criticism of growing economic disparities, and the aftermath of the global recession following the Lehman Shock made the continued pursuit of neo-liberalism difficult. Koizumi s successor was forced to find a new strategy, and Abe Shinzō appealed to the public with a conservative ideology (Nakakita 2014; Nakano 2015). In the end, the DPJ defeated the coalition government of the LDP and Komeito in the 2009 HR election, and rose to government. In order to secure a majority in the HC, the DPJ formed a coalition with the People's New Party, a new political party formed by defecting members of the LDP, and the Social Democratic Party (formally the Socialist Party). The DPJ resisted the clientelism and neo-liberalism of the LDP, and sought the realization of their policies of universal welfare, but were frustrated by a lack of financial resources. As internal divisions regarding fiscal reconstruction became increasingly fierce, many members left the DPJ and formed new political parties. Increasing public criticism of the DPJ government, and the absence of coordination between non-ldp political parties, allowed the LDP and Komeito to reclaim government at the 2012 HR election (Kushida and Lipscy eds. 2013; Maeda and Tsutsumi eds. 2015). The relationship between political parties was thus destabilized by the emergence of new parties, as well as by changes of government. Administrative and fiscal reform became the subject of debate, and the relationship between political parties and the state was re-assessed. There was an increase in non-aligned voters, and the electorate s support for the party system was shaken. Thus we can say that the de-institutionalization of the party system progressed during this period. Institutionalization of the LDP Under the multi-member SNTV system that used to be used in the HR elections, it was normal for 3 to 5 candidates to be elected from each electoral district. In order for large parties such as the LDP to maintain their seats in the Diet, they needed to mount multiple candidates in the same district. Following Leiserson (1968), we can understand the LDP governments of the past as coalitions of factions made up of individual Diet members: see Table 1 below. Different factions within the LDP would endeavor to gain party endorsement for candidates they had recruited, and continued to support the campaigns of those candidates even if they were not officially endorsed. 7 The LDP 7 Analyses of the relationship between the electoral system and factions include Cox and Rosenbluth (1996), Kohno (1997), Cox et al. (1999), and Krauss and Pekkanen (2011). On decentralized policy-making mechanisms in the LDP, see Sato and Matsuzaki (1986) and Krauss and Pekkanen (op. cit.). A comparative political science perspective on the LDP s factions is provided by 8

employed a system that enabled elected candidates to be recognized as LDP members regardless of whether or not they had received official endorsement, and this acted as a considerable limitation on the party executive s power to endorse candidates. The factions did not simply recruit candidates; they also exercised great influence on their promotion following election. Newly-elected Diet members would be affiliated with the faction that had supported them during their election campaign, and accumulate experience that would eventually lead to ministerial and other senior appointments. The recommendation of their faction was crucial to such appointments. Faction leaders would in turn rely on the support of their membership when they stood for election to party President. Furthermore it was not uncommon for presidential election procedures to be altered through negotiation among the factions (Tanaka 1986). Inter-factional struggles were intense, and reached their peak in the 1970s. There was even a case in which the Prime Minister was forced to dissolve the House of Representatives after a parliamentary vote of no confidence in the cabinet was passed due to the non-attendance of members affiliated with opposing factions. Under the multi-member SNTV system, it was difficult to differentiate among LDP candidates from within the same electoral district. Candidates thus cultivated a personal support base by protecting the interests of specific regions and/or industry sectors. This made clientelism the LDP s primary method of garnering votes, and its policy-making authority resultantly became diffused among members seeking to protect the interests of their own clientele. In addition, the high costs involved in individually-orchestrated election campaigns made candidates dependent on funds sourced from the factions. Opposition parties rarely nominated more than one candidate in any single district, and showed no desire to wrest power from the LDP. The low risk of actually losing government was a precondition for the development of factionalism and clientelism in the LDP. Split into factions, the LDP was forced even to engage in patronage demanded by support base beyond the party itself, and its decision-making focus was thus diffused. Once again employing the conceptualization of Randall and Svåsand (2002), in the internal and external structural dimensions of the LDP, the standard of institutionalization was not high. The interesting feature is that a highly institutionalized party system conversely created a dominant party that was exposed to the pressures of de-institutionalization. Under the new HR electoral system introduced in 1994, the LDP executive seized authority for candidate endorsements, and the power of its factions began to wane (Otake op.cit.; Takenaka 2006). Vote by party members took root as the means of nominating candidates and selecting party leaders (Uekami 2008; Smith and Tsutsumi 2014). Additionally, as the need for patronage in elections diminished, so too did the strength of the party s relationships with outside support groups seeking Boucek (2009). 9

favors (Köllner 2002). Furthermore, public subsidies to political parties came to represent 60% of the LDP s overall income, and 25% of income for individual Diet members, lowering the reliance on individual fund-raising. As a result, the role of factions in apportioning funds also diminished (Takenaka op.cit.; Carlson 2010). In other words, the standards of systemness and decisional autonomy in the LDP were raised, and degree of institutionalization on a structural level increased. In the internal and attitudinal dimension, clientelism was replaced by a conservative ideological pitch as the main strategy for garnering support. This attempt at value infusion is not inconsistent with the concept of institutionalization. In external-attitudinal terms, however, it must be noted that voter support has become more fluid. The onset of institutionalization could be said to be more prominent in the LDP internally. If the de-institutionalization of the party system is being accompanied by institutionalization of the dominant party, does this mean that the relationship between them is one of cause and effect? The next section brings a micro-level approach to bear on this question. 3. Change in Political Parties: Theories and Hypotheses The institutionalization of political parties and the party system are macro-level phenomena. In order to mount the argument that change in a party system prompts change in parties themselves, it is necessary to explicate the micro-level mechanisms of change. In this section we briefly review the existing literature on change in organization, policies, and governmental strategies of political parties, and present the theories and hypotheses employed in this article. Literature review How do changes take place in the organization, policies, and governmental strategies of political parties? There have been many attempts to verify theories and hypotheses in response to this question. Addressing these in detail is beyond the scope of this article; the following is simply a brief summary of salient points. Studies in the typology of political party organization have been undertaken in order to explain historical macro-phenomena, 8 with traditional theories such as spatial competition theory and coalition formation theory tending to treat each party as a unitary actor 9 (Downs 1957; Riker 1962; Kirchheimer 1966; Panebianco 1982). In response to this strand of research, Harmel and others have sought to explain specific changes at certain points in time by reference to factors internal to the parties concerned (Harmel and Janda 1994; Harmel et al. 1995; Harmel and Tan 2003). These studies look not only at external causes of change such as election losses, but also at how external stimuli 8 Krouwel (2006) provides a comprehensive review of typologies of political party organization. 9 Strom (op. cit., p.159) is critical of the view that conceives political parties as unitary actors. 10

are transformed into internal factors such as inter-factional power relations, power of the dominant factional coalition, and leadership of the party head, leading to organizational and policy changes. Such studies could be seen as seeking more precise explanations than those rendered by traditional political party research. A succession of studies in recent years has also noted the impact of intra-party politics on policy-making and participation in coalition governments. Some of the most recent findings include an account of inter-factional conflict and election loss leading to changes in a party s policies (Budge et al. 2010), and use of degree of internal division and presence/absence of a centralized system of power of as means of determining whether or not a party joins a coalition government (Ceron 2014). The approach of studies such as these is to explain party behavior by reference to internal factors. This article builds on these research developments by explaining political party organization and policy change with an emphasis on internal causes. We discuss how changes in the party system dictate the survivability of coalition governments and prospects for success in elections and, mediated by these factors, influence the power of internal factions and dominant factional coalition. Our approach in this article is not intended to deny that changes to the electoral system are a more fundamental independent variable influencing changes in party systems and even parties themselves (see footnote 7 in this regard). We do not employ such an account, however, as it cannot be used to explain specific changes taking place at particular points in time. Theory and hypotheses Harmel and Janda (op. cit.) state that a political party s performance criteria are dictated by the party s primary goal (p. 279, A5). In the analysis below, it is assumed that political parties seek votes and/or office in line with prevailing conditions. 10 The routes by which inter-party competition can influence internal party politics can be contemplated in terms of two different types of change: pattern and degree of inter-party competition. Firstly, change in the pattern of competition can surely lead to changes in a party s policies, because such change influences power struggles among different internal interests. H1: Change in the pattern of inter-party competition alters existing relationships between parties. This weakens the influence of dominant intra-party coalitions as monopoly holders of negotiation channels with other parties, and party policies change accordingly. Conceivable changes in the pattern of inter-party competition include a shift from single-party to 10 It is also theoretically conceivable that a party may have policy-seeking as its primary goal, but this possibility is not addressed in this article. There are two reasons for this omission. Firstly, in a single-member district system, vote-maximization is an effective means to gain power and achieve policy goals. Secondly, the LDP has traditionally been a vote-seeking party. 11

coalition government, and a reshuffle of partners in a coalition. In this article, such pattern changes as treated as external factors, and their influence on internal party politics is examined. When it becomes necessary to form a new coalition government, the pre-existing dominant factional alliance within the party may lose the partners it needed to manage the government, and other factions may emerge with connections to new partners, stripping the old coalition of its capacity for leadership both within and beyond the party. The influence of the old coalition may thereby diminish, or a new coalition may be formed, resulting in changes to the party s policies. This allows us to re-formulate the above hypothesis as a working hypothesis: H1a: Reorganization of coalition partners weakens the influence of the party s dominant factional alliance as monopoly holder of negotiation channels with other parties, and party policies change accordingly. Pattern change is not the only channel whereby inter-party competition can influence intra-party politics. Change in the degree of inter-party competition may affect the distribution of power within the party, resulting in changes to party organization. As inter-party competition grows more intense, it becomes essential for a party to formulate appropriate election strategies and approaches to day-to-day management astutely in response to other parties and public opinion trends. In order to maximize votes and seats in parliament, and to win or retain government, the party is required to achieve not only inter-factional compromise but also rational decision-making. The result, theoretically, is that powers such as candidate endorsement, personnel, and determination of governmental and party policies are entrusted to the party executive. Party organization becomes more centralized, and the influence of pre-existing internal groups such as factions grows weaker in relative terms. This centralization process can be understood also as a relationship between principal and agent. As principals, parliamentary members entrust some powers to their party executive the agent in order to achieve goals such as success in elections and appropriate management of government. 11 Accordingly, the influence exerted by pre-existing internal groups diminishes. This logical consequence can be summarized in the hypotheses below: H2: As inter-party competition grows, party organization becomes more centralized, and pre-existing internal groups are weakened. A slightly more concrete formulation is needed in order to test Hypothesis 2. The logic is sound, 11 See Ramseyer and Rosenbluth (1993) for a discussion of the LDP during the multiple-seat electoral system era. 12

but it is not necessarily clear in what circumstances the party executive would be strengthened and factional power diminished. One situation that may typically lead to enhancement of the party executive s leadership is victory in an election. Factional weakening is likely to result in an increase in the numbers of members not affiliated with any faction. This allows us to reformulate Hypothesis 2 as a working hypothesis: H2a: Election victory leads to an increase in the proportion of HR members not affiliated with any faction (non-affiliated members). Conversely, an election loss and resultant decline in leadership by the party executive is anticipated to cause an increase in the number of faction-affiliated HR members. This is because the relative rise in influence of the factions within the party means that factional affiliation becomes a more attractive proposition for members seeking to secure certain positions and achieve specific policy goals. Additionally, members choices in relation to factional affiliation are thought to be influenced not only by actual election results but also by the prospects of victory or defeat. This point will also be explored below. The incentive to join a faction is thought to diminish particularly for newly-elected members. If factional support is no longer necessary in order to gain party endorsement, there is no reason why a newly-elected member should join a faction after election. Furthermore, if the leadership of the party executive is strengthened as a result of victory in inter-party competition, there is likely to be especially little incentive for new members to join factions for purposes such as seeking higher office. Incumbent members, on the other hand, are as a rule already entitled to endorsement, so have no need to become obligated to the party executive. Many veteran LDP members in office since the pre-reform heyday of the factions would have gained endorsement and achieved promotions by affiliating themselves with factions. New non-affiliated members provide the foundations of power for the party executive, but they do not furnish their support unconditionally. Generally, new members have a weak support base within their constituencies: election outcomes are thought to be swayed to a large degree by the popularity of party leaders. As long as the party leader remains popular, therefore, non-affiliated members are sure to support the party executive. These ideas are encapsulated in the following working hypotheses: H2b: In general, new HR members tend not to join factions. H2c: New HR members tend not to join factions if their party wins an election. H2d: New non-affiliated HR members support popular party leaders, but do not support unpopular party leaders. 13

As regards Hypothesis 2b, in order to establish the tendency for new members not to join factions, it is necessary to pursue comparison with members elected several times. For Hypothesis 2c, comparison between successful and unsuccessful elections must be used to establish whether or not the tendency not to join becomes more pronounced. Hypothesis 2d must be tested by establishing the relationship between leader popularity and support through comparison with new HR members that are affiliated with factions. It might be envisaged that both the approach to factions and the level of support for the party leader displayed by new members will vary depending on whether or not the member is affiliated with a faction forming part of the dominant coalition. 4. Testing the Hypotheses This section tests the hypotheses outlined above. Firstly, however, it provides an overview of the factions that have exerted major influence on the institutionalization of the LDP, and their respective policies. Table 1 shows the number and proportions of House of Representatives (HR) members affiliated with the LDP s different factions since the 1990s. Factions, which had been part of the LDP ever since the party s formation, coalesced in the early 1970s into four major factions and one smaller faction, namely the Heisei Study Group, Kōchi Group, and Seiwa Policy Study Group, as well as the pre-existing Policy Science Group and Banmachi Policy Study Group. For the purpose of convenience, in this article the labels Heisei, Kōchi, Seiwa, Others, and Non-affiliated are used. The Table also shows the names of the faction leaders and which factions produced the party leaders (grey shaded cells). The five-faction era lasted many years, but divisions began to appear in the 1990s. At the start of that decade the Takeshita faction split in two groups, with one of the two groups, the Hata faction, leaving the LDP and forming a new party. Meanwhile, the traditional Kōchi Group suffered the departure of a breakaway group led by Kōno Yōhei, and a rift developed within the remaining membership concerning the faction s management (this was finally resolved in 2008). The Watanabe faction split into the Yamasaki faction and Former Watanabe factions, and the latter combined with the Nakayama/Kamei faction, a splinter group from the Mitsuzuka faction, to form a new faction, the Shisui Group. Table 1. Factional structure of the LDP (House of Representatives members only) 14

April, 1990 Heisei Kōchi Seiwa Takeshita Miyazawa Abe 22.1% 20.0% 20.4% (63) (57) (58) Policy science Nakasone 15.8% (45) Others Banmachi Non-affiliated Kōmoto othres 8.1% 1.1% 12.6% (23) (3) (36) March, 1993 July, 1993 March, 1994 October, 1996 August, 1997 February, 1999 June, 2000 February, 2001 November, 2003 February, 2004 September, 2005 February, 2006 Hata Obuchi Miyazawa Mitsuzuka Nakasone Kōmoto 11.4% 9.5% 19.0% 18.7% 13.6% 8.1% 2.6% 17.2% (31) (26) (52) (51) (37) (22) (7) (47) General Election Obuchi Miyazawa Mitsuzuka Watanabe Kōmoto 14.0% 24.4% 24.4% 19.9% 9.5% 0.0% 7.7% (31) (54) (54) (44) (21) (0) (17) General Election Obuchi Miyazawa Mitsuzuka Watanabe Kōmoto 19.2% 22.4% 25.3% 19.6% 6.9% 0.0% 6.5% (47) (55) (62) (48) (17) (0) (16) Obuchi Katō Mori Nakayama/Kamei Old Watanabe Yamasaki Kōmoto 20.5% 18.6% 16.0% 6.8% 9.5% 10.6% 6.1% 0.0% 11.8% (54) (49) (42) (18) (25) (28) (16) (0) (31) General Election Hashimoto Horiuchi Katō Kōno Mori Etō/Kamei Yamasaki Kōmura 25.4% 13.6% 5.1% 4.7% 16.9% 14.8% 8.1% 5.1% 1.7% 4.7% (60) (32) (12) (11) (40) (35) (19) (12) (4) (11) General Election Hashimoto Koga Ozato Kōno Mori Kamei Yamasaki Kōmura 21.0% 12.8% 4.5% 3.7% 19.8% 12.3% 9.9% 4.9% 1.6% 9.5% (51) (31) (11) (9) (48) (30) (24) (12) (4) (23) General Election Tsushima Koga Tanigaki Kōno Mori Ibuki Yamasaki Kōmura 12.3% 10.6% 3.8% 3.4% 20.5% 6.2% 10.6% 3.8% 3.8% 25.0% (36) (31) (11) (10) (60) (18) (31) (11) (11) (73) January, 2009 August, 2009 February, 2010 December, 2012 February, 2013 Tsushima Koga Asō Machimura Ibuki Yamasaki Kōmura 14.8% 16.4% 4.9% 20.1% 7.2% 12.5% 4.6% 4.6% 14.8% (45) (50) (15) (61) (22) (38) (14) (14) (45) Nukaga Koga Asō Machimura Ibuki Yamasaki Ōshima 11.0% 19.5% 6.8% 16.1% 8.5% 11.9% 1.7% 0.0% 24.6% (13) (23) (8) (19) (10) (14) (2) (0) (29) Nukaga Kishida Asō Machimura Nikai Ishihara Ōshima 9.7% 10.4% 8.7% 18.0% 7.3% 3.8% 2.4% 0.0% 39.8% (28) (30) (25) (52) (21) (11) (7) (0) (115) Note: Percentages show the proportion of all LDP HR members that are affiliated with the faction. The actual number of such members is shown in parentheses. Grey shading indicates the LDP leader's faction. Source: Kokkai-binran General Election General Election Contrary to the impression given by their official names, factions of the LDP could not be characterized as policy groups: they are groups founded on personal connections among individuals. Nevertheless, as the political convictions of a faction s founder are internalized by its followers, ideological differences can emerge across different groups (Zakowski 2011, p. 181). We therefore measured the overall policy positions of each faction by aggregating the policy positions of HR members affiliated with each faction. Data for this purpose was taken from The UTokyo-Asahi Survey (UTAS). 12 This survey has asked all candidates in national elections since 2003 to state their 12 UTAS was conducted by Taniguchi Masaki of the Graduate Schools for Law and Politics, the University of Tokyo and the Asahi Shimbun. For further details, refer to the UTAS website: 15

positions on a range of issues. Response rates are in excess of 90% in each survey, enabling it to measure the policy positions of almost all HR members. UTAS does not include data from the 1990s, but the membership of factions has not changed markedly since then, meaning that it is possible to use the survey results to analogize policy positions in the 1990s. Principal component analysis was used to summarize HR members positions on a range of policy issues in accordance with two dimensions: liberal/conservative and big government/small government (see Appendix). Mean principle component scores and standard deviation for members of each faction were calculated as shown in Table 2. Table 2. Factional policy positions Liberal (-) - Conservative (+) Big gov. (-) - Small gov. (+) 2003 2005 2009 2012 2003 2005 2009 2012 Heisei.333.444.614 1.026 -.160.663 -.982 -.711 (0.460) (0.440) (0.551) (0.496) (0.938) (0.969) (0.539) (0.753) 46 30 13 26 46 30 13 26 Kōchi.252.161.383.799.031.876-1.089 -.705 (0.568) (0.578) (0.544) (0.480) (0.824) (1.000) (0.571) (0.666) 36 39 23 25 36 39 23 25 Seiwa.537.605.942.981.367.821 -.823 -.725 (0.537) (0.544) (0.531) (0.440) (0.918) (0.837) (0.493) (0.718) 41 52 19 51 41 52 19 51 Others.440.502.606.917 -.033.533 -.856 -.724 (0.583) (0.429) (0.486) (0.471) (0.883) (0.724) (0.592) (0.825) 67 67 33 63 67 67 33 63 Non-affiliated.259.457.524.963.373.865 -.694 -.573 (0.440) (0.578) (0.464) (0.390) (0.621) (1.036) (0.844) (0.788) 21 85 24 107 21 85 24 107 Total.386.452.600.947.068.755 -.878 -.662 (0.538) (0.537) (0.530) (0.438) (0.886) (0.922) (0.635) (0.768) 211 273 112 272 211 273 112 272 Note: The number shown in upper row is the mean principal component score of the faction members. The number shown in middle row indicates the standard deviation, and that shown in lower column indicates the number of members of the faction. In the liberal/conservative dimension, factional policy positions became gradually more conservative between 2003 and 2012. Analysis of policy positions in the big government/small government dimension shows that in 2005 small government was favored in line with the position LDP leader Koizumi Jun ichirō, but that the position shifted to big government in 2012, when initiatives such as monetary easing and public works were proposed under the banner of Abenomics. There is also some degree of variation in positions across the different factions. Over the period surveyed the LDP as a whole became increasingly conservative, but the tendency to http://www.masaki.j.u-tokyo.ac.jp/utas/utasindex.html. Data used in this study is available on this site. 16

conservatism was particularly strong in the Seiwa group, while the Kōchi group maintained a relatively liberal position. 13 In terms of positions on government size, the Heisei Group previously took a slightly more pro-big government position than other factions, but this difference had virtually disappeared by 2012. Testing the hypotheses: Change in the pattern of inter-party competition Firstly, we test the hypothesis concerning the mechanism by which change in coalition partners renders LDP policy more conservative. H1a: Reorganization of coalition partners weakens the influence of the party s dominant factional alliance as monopoly holder of negotiation channels with other parties, and party policies change accordingly. Earlier we explained that after losing government in the 1993 election, the LDP formed a new coalition government with the Socialist Party and New Party Sakigake. Katō Kōichi, the former Chairman of the Policy Affairs Research Council and the Secretary-General who led the management of the coalition within the LDP, was originally from the Kōchi group, while his Deputy Secretary-General Nonaka Hiromu was from the Heisei group. These factions were the traditional mainstream factions forming the dominant coalition within the LDP, and were known for taking a relatively liberal policy stance in comparison with other LDP factions. In order to maintain the coalition with the Socialists in particular, it was essential for the LDP to allow liberal policies to prevail. This coalition government could therefore be said to have underpinned the dominance of the liberal factions within the LDP (Nakakita 2014). Some LDP members began to call for policy discussions to be initiated with the New Frontier Party, which was more conservative than the Socialists and closer to the LDP on the policy spectrum. A power struggle broke out between Katō and others who favored maintaining the coalition government with the Socialists and Sakigake on the one hand, and conservative members calling for change on the other. Ultimately, the Socialist Party and Sakigake suffered heavy losses in the 1996 House of Representatives and 1998 House of Councilors elections, and their coalition with the LDP dissolved. The Democratic Party of Japan (DPJ) subsequently emerged to replace the Socialist Party and Sakigake as the main contender for regime change. The LDP also lost seats and its majority in the HC in the 1998 election. Prime Minister Hashimoto Ryūtarō resigned, and Katō Kōichi, seen as the leader of the liberals within the LDP, also relinquished his post as Secretary-General. New Secretary-General Mori Yoshirō was a member of the Seiwa group, known as the LDP s conservative faction. In 1999, the LDP formed a new coalition 13 See Zakowski op.cit. for information on the liberal stance of the Kōchi group. 17

government with the Liberal Party and Komeito. Nonaka became opposed to Katō on the issue of which partners to include in this coalition, and the dominant factional alliance within the LDP was thus altered. A new dominant factional coalition was formed by the Heisei group together with the more conservative Seiwa and Shisui groups. Nakakita explains that the process of formation of the new coalition government brought an end to the predominance of liberal factions within the LDP, and enabled a new conservative ascendancy. As this shows, Hypothesis H1a is supported. Testing the hypotheses: Change in the degree of inter-party competition Here we test the hypothesis concerning the mechanisms whereby heightened competition among parties leads to a decline in factionalism within the LDP. H2a: Election victory leads to an increase in the proportion of members not affiliated with any faction (non-affiliated members). Under the 1955 system, the great majority of LDP Diet members had factional affiliations. Table 1 shows that this tendency continued almost unabated into the 1990s. In the 2000s, however, there was an increase in the proportion of HR members not affiliated with any faction. What can be concluded regarding the timing at which such non-affiliated party members increased? The increase followed the 2005 HR election which was called after Prime Minister Koizumi s postal privatization bill had been defeated. At this election, despite the LDP s refusal to endorse incumbent candidates who had opposed the bill and choice to nominate alternative candidates instead, the party still achieved a crushing victory, claiming 296 seats. The proportion of non-affiliated HR members jumped to 25%. Following Koizumi s replacement by new Prime Minister Abe Shinzō, however, a series of cabinet-level scandals saw approval ratings plummet, and the LDP suffered major losses in the 2007 HC election. Abe resigned and was replaced by Fukuda Yasuo, and then Asō Tarō, but support for the LDP continued to languish. The party was (as expected) defeated soundly at the 2009 HR election, losing government to a coalition of parties led by the DPJ. By this point, the proportion of non-affiliated HR members in the LDP had fallen to 15%. The DPJ thus seized power from the LDP in 2009, but voter support declined steadily in the following years; by the time the next HR election was held in late 2012, another regime change was seen as inevitable. The proportion of non-affiliated HR members in the LDP had risen again to one quarter after the 2009 defeat, and by August 2012, around one-third of members had no factional affiliations. This proportion grew further to around 40% after the LDP regained power following a major victory in the 2012 HR election. Hypothesis 2a predicted that the proportion of non-affiliated members would rise following an election victory or prospect thereof. This tendency has been observed since the election strategies of 18