Gary W. Cox

Department of Political Science

University of California, San Diego

La Jolla, CA 92093-0521

Frances M. Rosenbluth

Department of Political Science

Yale University

New Haven, CT 06520-8301

Michael F. Thies

Department of Political Science

University of California, Los Angeles

Los Angeles, CA 90095-1472

March 1998

In this paper, we explore the link between electoral rules and the factionalization of Japan's long-dominant Liberal Democratic Party by taking advantage of the "natural experiment" that the Japanese Upper House provides. After describing the three different methods used to elect members of the Upper House over the period studied, we articulate several hypotheses: some deriving from the different electoral incentives faced by Lower-House candidates and the three groups of Upper House candidates; some deriving from the presence of an alternative set of office incentives (progressive ambition); and one deriving from factional bosses’ incentives. Consistent with our hypotheses, we find that the rate of factional affiliation is highest among Lower House members and post-1983 Upper House national-district members, lower among Upper House prefectural members, and lowest among pre-1983 Upper House national-district members. We also find that smaller factions died out sooner in the Upper House, and that the distribution of factional strengths was more skewed there. Finally, although the prospect of moving to the Lower House does not affect the probability that an Upper House member will join a faction, we find Upper House members with cabinet experience are more likely to be factionally affiliated than those without.

The literature on party systems contains two central themes. The first is that the number and characteristics of parties can be traced to the incentives produced by electoral rules. The best known hypothesis in this tradition is "Duverger’s Law," which posits that "the simple-majority single-ballot system favors the two-party system" (Duverger 1954). The second tradition holds that party systems emerge as by-products of social cleavages, as the political manifestation of primordial divisions within societies. The seminal work in this latter tradition is Lipset and Rokkan’s "Cleavage Structures, Party Systems, and Voter Alignments" (Lipset and Rokkan 1967). More recently, scholars have brought the insights of the two strains in the literature together. Thus, Riker (1982) refines Duverger’s Law with explicit attention to pre-existing cleavage structures, while Ordeshook and Shvetsova (1994) and Amorim-Neto and Cox (1998) show how electoral rules and cleavage structures interact in generating and sustaining political parties.

Although the most recent analyses do attempt to control for cleavage structures in assessing the impact of electoral rules, they can do so only with relatively crude indicators (such as the effective number of ethnic groups). There is thus ample room to worry that these studies have not fully controlled for the nature and nuance of cleavages in the range of polities they examine. Fortunately, some countries produce "natural experiments" that provide better means of controlling for differing cleavage structures, due to their possession of bicameral legislatures. Concurrent, or close-to-concurrent elections for two separate legislative chambers in the same country better allow us to see the impact of electoral rules, since social cleavages—and other country-specific factors, such as political culture—are held constant. Of course, if both houses are elected under the same rules, then there will be no reason to expect any differences due to electoral incentives. But the more "incongruent" (Lijphart 1984) a country’s bicameralism is—i.e., the more dissimilar are the methods of election used to fill seats in the two legislative chambers—the more likely are differences in the party system.

Candidates and parties should adapt to the particular electoral niches created by the rules for each house and, insofar as those niches differ across chambers, the number, size, and internal organization of parties should differ as well. For example, a more "permissive" electoral system in one chamber—one with a lower threshold of representation (Lijphart 1984)—might sustain parties that cannot hope to win in the other chamber, with its less permissive rules. Cox (1997:19-27), in a study of sixteen bicameral systems, shows that differences in the party systems of upper and lower houses do tend to follow differences in their electoral rules.

In this paper, we push the comparison between upper and lower houses a step further, examining variation in the numbers and characteristics of intraparty structures, better known as party factions. We focus on Japanese politics, where factions have been the focus of a great deal of scholarship but where, again, little attention has been paid to the potential explanatory leverage provided by bicameralism. If different electoral rules can create different incentives for the formation and sustenance of intraparty factions—a key contention of many recent studies of Japanese elections—then intercameral differences in electoral rules ought to produce different factional structures in the two chambers.

The remainder of the paper proceeds as follows. In section 2, we review the literature on Lower House factions, and concentrate specifically on the theory and evidence supporting the notion that factions were a response to the incentives inherent in the SNTV electoral system. In section 3, we explore the incentives facing Upper House LDP members and the role that Lower-House-style factions could and could not play in helping members respond to those incentives. In section 4, we answer the questions we have raised about inter- and intra-chamber differences in factionalization. To do this, we have amassed a data set on the factional affiliations of all Upper House LDP members -- both in districts and in the national constituency -- from 1962 to 1992. To our knowledge, this is the first time such data have been gathered and analyzed in this fashion. Section 5 concludes.

Many students of Japanese politics argue that the peculiar single non-transferable vote (SNTV) system formerly used in elections to the Japanese Lower House promoted the factionalization of the long-ruling Liberal Democratic Party (LDP). The argument, in brief, goes as follows. First, the number of seats per electoral district under the old Lower House system hovered around four, which meant that any party seeking a majority of seats had to win an average of at least two seats per district, hence had to run an average of at least two candidates per district. Second, because citizens cast a single vote (for an individual candidate), a majority-seeking party’s candidates in each district were unavoidably put into direct competition with one another for votes. The system has sometimes been described as combining what in the U.S. context are separate primary and general elections. Third, the rank and file’s demand for a competitive advantage over their copartisan rivals was met by factional bosses, who helped members get three crucial aids to re-election: the party endorsement, financial backing, and party and governmental posts. In return, the boss received his followers’ support in the LDP presidential election, which he could use either to pursue the party presidency (hence, the premiership of Japan) himself or to trade for other positions. SNTV thus engendered systematic and organized intraparty competition, with long-standing factions supplying rival candidates with campaign and career support in the roughly 130 Lower House districts.

We highlight here two consequences of the SNTV electoral rules for the Lower House’s particular variant of factionalism. First, while factional bosses may have wanted to recruit as many candidates into their camps as possible, they usually had to settle for at most one candidate from any given district. Candidates joined factions, after all, to distinguish themselves from other LDP candidates. This limit on the number of their members per district, however, naturally limited the size of Lower House factions.

Second, factional bosses had to adjudicate the competing claims of their members for the most desirable party and government posts. A seniority system for post allocation, which became an integral part of the factional system in the Lower House, gave back benchers enough security to stay in a single faction over the length of their careers (Sato and Matsuzaki 1986; Kohno 1992; Epstein, Brady, Kawato, and O’Halloran 1997).

If electoral incentives were indeed important in conditioning Japanese factionalism, then what did they look like in the Upper House? During the period under study here, Members of the House of Councillors (MHCs) were not elected under the same electoral rules as were their counterparts in the House of Representatives. But they were exposed, presumably, to the same cultural currents and cross-currents. Thus, the Upper House offers something of a natural experiment in the study of factionalism. The different electoral incentives faced by MHCs and MHRs suggest that factionalism should have looked systematically different in the two houses; the similarity of the cultural incentives suggests that factionalism should have been much the same. In the next section, we articulate just what should be different, from an electoral-incentives perspective. In order to do this, we first describe the rules of election in the Upper House.

Since 1983, Upper House members have been elected in two main ways. First, 100 members are elected (50 every third year) under a proportional representation rule from closed party lists in a nationwide district. Second, the remaining 152 members are elected (76 every third year) from Japan’s 47 prefectures by the SNTV rule. Prior to 1983, Upper House elections were the same in the prefectural districts but the 50 members to be selected every third year from the national district were elected by SNTV, rather than by closed-list PR. Thus, all told, there have been three different electoral systems used in the Upper House over the period we study: nationwide SNTV (1962-83); nationwide closed-list PR (1983-92); and prefectural SNTV (throughout). We shall consider in turn the electoral incentives to join factions that MHCs elected under each of these three distinct sets of rules should have had. We then consider other incentives that they might have had, those not related to their desire to win or retain a seat in the Upper House.

The prefectural MHCs. Since most of the prefectural districts in the Upper House returned only a few members, and since only half the total number was elected at any given election, direct competition between LDP candidates in general elections was much rarer in the Upper than in the Lower House. In the 1995 election, for example, 24 of the 47 prefectures elected one representative each, 18 elected two members, 4 elected three members, and 1 chose four. Intraparty competition in the single-member districts arose only when the party could not agree on a single nominee (about 10% of the cases) and was not much more likely in the two-seat districts. Thus, most of the LDP’s Upper House prefectural candidates – 62% over the period we study – did not face intraparty competition and hence did not face the usual incentives (cited in the literature on Lower House elections) to seek out factional help with their campaigns. Indeed, Watanabe (1964:191) suggests that, because Upper House districts encompass several Lower House ones, many Upper House candidates form alliances with all correspon/ding Lower House incumbents. These alliances cannot be restricted to a single faction unless all the Lower House incumbents in a prefecture are from the same faction (which has rarely occurred).

The national MHCs under SNTV. What of the national candidates elected by SNTV before 1983? They did face intra-party competition, but they ran for election in the nation at large. Factions, geared to the local elections held in the Lower House, had much less to offer them; a factional donation of money comparable to that given to Lower House candidates would have mattered little in a national campaign. This is particularly true given that television and radio advertising, which would have been the most efficient way to campaign before a national constituency, were banned by law. Thus, the strong candidates were either celebrities -- actors, news anchors, popular singers and sports stars who already had national reputations, and who therefore had a huge comparative advantage in terms of name recognition (Curtis 1976: 61-64) -- or representatives of nationally organized interest groups (Watanabe 1964: 191). Neither celebrities nor interest groups needed the factions for purposes of getting elected to the Upper House.

The national MHCs under list-PR. After the national district electoral rules were changed to introduce closed party lists, intraparty competition for votes in the national district disappeared altogether -- voters now could vote only for a party slate, without any ability to choose among a party’s candidates. However, this change should have strengthened rather than weakened the role that factions could play. Under the closed list regime, whether one got elected or not depended almost entirely on whether one got a good list position, and the process of deciding who got what list positions could be brought into the usual factional bazaar that decided Lower House nominations (on which, see Cox and Rosenbluth 1996). It is true that the party, troubled by the national list’s poor performance, tried to inject more competition for spots on the list. The Election Strategy Committee now requires that, to get on the list, a candidate first bring 20,000 new rank and file members into the party and document the existence of one million members in his or her koenkai (Kato 1987:120-121). In addition, the LDP has tried to present the entire list to the public under a united party label, with slogans such as "Security, Safely, Stability" and "A Vigorous Japan, Vigorous Politics" (Ibid.). Nonetheless, even after these reforms (which were not too successful in boosting the list’s performance), Iseri (1988:44) writes that "the major concern for candidates running in the PR district is to secure a good position on the party list. This depends not only on how many rank-and-file members the candidate ‘brings into’ the party, but also on whether he is supported by a faction."

Contrasts with the Lower House. What differences, then, do we expect in the nature of factions in the House of Councillors, as opposed to those in the House of Representatives? First, in the Lower House virtually everyone faced strong electoral incentives to join factions -- so we expect almost everyone to join a faction. In contrast, the reasons for joining factions in the Upper House were more varied and generally weaker. Prefectural candidates typically did not face intraparty competition, so factions were less often in a position to be useful. National candidates under SNTV were even less likely to benefit from factional affiliation, because they had to compete at large in the nation. Only national candidates competing under the list-PR system that came in after 1983 had clear incentives to join factions. Hence, we expect that the proportion of members affiliated with factions should have been highest among MHRs and among national district (list) MHCs; lower among prefectural district MHCs; and lower still among national district (SNTV) MHCs. To keep things straight, we shall label this expectation Contrast Hypothesis 1, and label the hypotheses below in a similar fashion.

Second, part of the mechanism stabilizing the factional bargain in the Lower House was the well-entrenched seniority system noted above, which served largely to regulate access to the plum posts in the cabinet. This system gave Lower House members a strong incentive to join a faction and stay in that faction, in order to get into the cabinet as quickly as possible. Indeed, all MHRs who followed this strategy of joining and staying did get their turn in the cabinet, if they succeeded in winning reelection the requisite number of times (cf. Sato and Matsuzaki 1986). In contrast, the number of cabinet posts allocated to MHCs was much smaller relative to the total conservative membership. Thus, relatively few MHCs could realistically aspire to cabinet posts, and their incentives to join and stay in a particular faction, waiting patiently in the seniority queue, were accordingly smaller. Therefore, at least before the introduction of list PR in 1983, one expects that relatively more MHCs than MHRs will spend some portion of their careers without any factional affiliation (Contrast Hypothesis 2).

Of course, the rule change for election to the national district did not simply create a new class of parliamentarians from whole cloth. Incumbents, elected in 1977 or 1980, if they were interested in reelection, had to run in 1983 or 1986 (respectively) under the new rules. The new rules made list position important, and that was controlled by faction leaders. From this follows Contrast Hypotheses 3, that incumbents elected in the national district without factional affiliation before the 1983 rule change should have joined factions in their pursuit of reelection after the rule change.

Fourth, as we noted above, the strong electoral incentives facing MHRs regulated the size of factions in the Lower House. In the Upper House prefectural districts, in contrast, a single faction should have been able to run two candidates from a single two-seat Upper House district, for example, because the two seats were filled at different times (as in U.S. Senate races). Thus, the potential for internecine competition between the two members was removed and the upper bound on factional size implied by the Lower House’s electoral rules was also removed. Consequently, comparing prefectural candidates in the Upper House to their Lower House counterparts, we expect that having two MHCs from the same district and faction serving at the same time will not be as unusual as having two MHRs from the same district and faction serving at the same time (Contrast Hypothesis 4a); and we consequently expect a more skewed distribution of factional sizes in the Upper House than in the Lower House (Contrast Hypothesis 4b).

Progressive ambition. Even though the "usual electoral incentives" less frequently obtained in the Upper House, this does not mean that factions should have been entirely absent from the Upper House. It might have meant this if the Upper House had been hermetically sealed off from the rest of the political system. But many members of the Upper House had their eye on continuing their political career elsewhere -- running for a Lower House seat, running for a seat in their prefecture’s assembly, running for their prefecture’s governorship, landing a spot in the cabinet, and so on. Thus, whereas analysts typically assume that members of the Lower House in Japan wish to continue in the Lower House, the appropriate assumption for many Upper House members was (and is) that they wish to continue their careers elsewhere, at least eventually. In terms of the well-known distinction made by Schlesinger (1966), members of the Lower House can be characterized as having static ambition (the desire to continue where they are) while many (if not all) members of the Upper House can be characterized as having progressive ambition (the desire to move up in the hierarchy of posts).

Those MHCs who did have progressive ambitions in the period we study would presumably do what it took to further those ambitions. If affiliating with a faction was useful in angling for a post in the cabinet, then MHCs who fancied themselves ministerial timber would have had an incentive to join a faction. If affiliating with a faction was useful in angling for a nomination to the Lower House, then MHCs who did wish to move on to the House of Representatives would have had an incentive to join a faction. If affiliating with a faction was useful in angling for the LDP’s nomination as prefectural governor, then MHCs who did wish to become governors would have had an incentive to join a faction. All of these incentives stemmed from members’ perceptions of what it would take to secure "higher" office, not from their perceptions of what it would take to retain their seat in the House of Councillors. They were incentives stemming from progressive rather than static ambition.

If these incentives to pursue higher office were in fact important, we should be able to find some evidence of it. First, one expects MHCs who at some point landed posts in the cabinet to have higher factionalization rates than those who never did, other things equal (Progressive Hypothesis 1). To the extent that the factions guarded access to the cabinet jealously, which they appear to have done, part of the price for admission would generally have been joining some faction.

Second, one might expect MHCs who at some point sought seats in the House of Representatives to have higher factionalization rates than those who did not. The reason for the "might" has to do with the typical strategy described by Watanabe (1964:191), in which Upper House candidates sought to form alliances with all of the Lower House candidates included in their prefecture. If a politician planned to seek a Lower House seat in the near or medium term, should she immediately join a faction, or should she wait until the moment at which she launched her Lower House campaign? The first strategy, of joining during one’s Upper House career, makes it harder to get help from the full array of Lower House candidates in one’s prefecture, especially from those who suspect that one will run against them in the near future. Thus, only if joining a faction early helps one secure that faction’s support for a later run might there be a clear incentive deriving from this form of progressive ambition to join during one’s Upper House career. But anecdotal evidence suggests that factions chose candidates mostly on the basis of their anticipated ability to win elections, not on the basis of the length of their affiliation with the faction. So, on balance, it is not obvious that there would be an incentive to join a faction early in order to position oneself for an eventual run at the House of Representatives. There is no clear theoretical prediction (although we will look into the matter anyway, as Progressive Hypothesis 2).

MHCs as targets of factional leaders’ progressive ambitions. While MHCs might have faced fewer electoral incentives to form or join factions than their MHR counterparts, the factions had ample incentive to woo MHCs once they took office. One of the best known roles that factions played within the LDP was to organize the selection of the party president (who, by virtue of the party’s perpetual majority, automatically became prime minister). Factional leaders become factional leaders in order to vie for the top job. When a factional leader threw his hat into the ring, his followers were expected to line up in support of his candidacy; when the leader chose not to run, his followers were expected to support whichever candidate the leader endorsed. This was the quid pro quo for the help the leader had provided them throughout their careers. In order to better pursue the LDP presidency, factional bosses had an incentive to increase the size of their factions. As we have said, there is a natural ceiling on the size of Lower House factions -- no faction can run more than one candidate in a given constituency very often, so no faction is likely to include more Lower House members than there are Lower House districts. Moreover, even attaining this size is extremely difficult because it entails being competitive in every district.

But the party constitution provides every member of the party from both houses with one vote in party presidential elections. MHC votes are just as valuable as MHR votes, and adding MHCs to a faction’s roster does not threaten the career prospects of its MHRs. Thus, factions have an incentive to "colonize" the Upper House (Watanabe 1966:229; Honzawa 1990:199ff, 259ff). For this reason, we expect factional representation in the Upper House to be skewed in favor of the biggest factions, those whose leaders have a real shot at the party presidency (and for whom the marginal vote is most valuable) (Target Hypothesis 1).

Static ambition redux. While most MHCs did not seek to carve out careers in the House of Councillors and did not face the sort of intraparty competition that their Lower House colleagues typically did, there were nonetheless some who wished to be reelected and who did face intraparty competition. These members had, above and beyond any other incentives they may have faced to join factions, a specific and short-term incentive of the usual sort. Thus, other things equal, we expect those MHCs who did face intraparty competition at the general election to have affiliated more often with a faction (Static Hypothesis 1).

To examine the attributes of Japanese Upper House factions, we compiled a dataset of all conservative winners between 1962 and 1992. Sources of information about factional affiliation included the election coverage of major national newspapers throughout the period under examination (Asahi shimbun various issues; Yomiuri shimbun various issues), as well as "Who’s Who" type directories (Kokkai benran various issues; Seiji handobukku various issues). We also compiled a complete listing of all cabinet ministers, and their partisan and factional ties, drawing heavily on Sato and Matsuzaki’s Jiminto seiken [The LDP Administration], and the newspapers and rosters mentioned above. Finally, comparison to Lower House factions was facilitated by Reed’s Japanese Electoral Data, and our own additions and corrections to all of the above sources. To our knowledge, this is the first time the complete Upper House electoral data and cabinet data have been compiled and used for such an analysis.

In this section, we present evidence that bears on the various hypotheses sketched out in section 3. We proceed sequentially, labeling each paragraph with the name of the hypothesis being examined.

Contrast H1: Figure 1 displays the inverse factionalization rates of each of three groups of conservative Diet members: MHRs, MHCs from the prefectural districts, and MHCs from the national district. The horizontal axis gives the years from 1960 to 1993, while the vertical axis gives the percent without known factional affiliation. As can be seen, the lowest line is that corresponding to MHRs, the second lowest line is that corresponding to prefectural MHCs, and the highest line (up until 1980) is that corresponding to national MHCs. The national MHCs’ line drops dramatically in 1983 (the first election using party-list PR) and is as low as the line for MHRs in 1983-93.

These patterns are consistent with our expectations. Lower House members and Upper House national members competing under list rules have the greatest incentives to join factions, Upper House candidates from the prefectures have intermediate incentives, and Upper House national members competing under SNTV rules have the least. Thus, the lines are ordered as expected and, in particular, the large and abrupt drop in the percent without factions among national MHCs appears to be explained by the shift in electoral rules from SNTV to list PR.

Contrast H2: Much the same patterns are evident if one looks at individual politicians over their careers, instead of aggregating within years (see Table 1). Of 77 Upper House national winners whose first win came before the introduction of the list system (1965-80) -- and who thus had the weakest incentives to join factions -- 42.9% started their careers without any factional affiliation and 15.6% never acquired such an affiliation. Moving to the group with the next strongest incentives, of 294 Upper House prefectural winners in the period 1962-92, 23.1% started their Upper House careers without any factional affiliation and 13.3% never acquired such an affiliation. In contrast to these figures, of 832 Lower House winners in the period 1960-93, only 12.3% started their Lower House careers without any factional affiliation and only 2.2% never acquired such an affiliation. Similarly, of 51 Upper House national winners whose first win came after the introduction of the list system (1983-95), 5.9% started their careers without any factional affiliation and 5.9% never acquired such an affiliation.

Contrast H3: What about Upper House national winners whose careers spanned the 1983 rule change from SNTV to closed-list proportional representation? If the new electoral system was more conducive to factionalization, then we should find that any holdovers joined factions in order to adapt to the new system. Of course, a focus only on members who won both before and after the rule change will ignore members who might have failed to receive renomination, perhaps due to their reluctance to join factions, but in that sense, the bias is against our finding any evidence of rational adaptation. As it turns out, only thirteen LDP Upper House members were successful under both systems. Of those, two were already factionally affiliated before 1983, while eleven were unaffiliated. And here the results are the strongest yet that electoral rules matter for factions: all eleven previously unaffiliated members saw fit to join factions in order to be renominated and reelected in 1983 or 1986. Interestingly, both of the members who had already been affilated with a faction chose to switch to another faction (from the Fukuda faction to the Suzuki faction in both cases) at the very same time. We are not sure why the two made the change, but an explanation consistent with the electoral incentives theory would be that they saw their chances of securing a good spot on the party list as better in their new faction than in their old one, perhaps for reasons of proportional access to list slots across factions.

Contrast H4a: Table 2 contrasts the frequency of "double returns" for factions from single districts in the Upper and Lower Houses. There were 115 pairs of Upper House prefectural winners from the same district and faction serving staggered terms over the period studied (1962-92). There were also 2 cases (out of 43 cases with two conservative winners) in which both winners in a given year and district were from the same faction. Thus, all told, there were 117 cases of "two conservatives from the same faction serving in the same district at the same time (not necessarily elected at the same election)." In contrast, the corresponding figure for the Lower House -- in which staggered terms were not possible -- was 46. Given 11 years ´ 46 districts in the upper house, these figures mean that on average about 23% of the Upper House prefectural districts had two conservatives from the same faction after an election. In contrast, given 13 years ´ 129 districts in the lower house, on average only 3% of the Lower House districts had two conservatives from the same faction after an election.

Contrast H4b: Given that there was no problem in returning two members from a given prefecture, was the largest factional contingent in the House of Councillors larger than that in the House of Representatives? It is already well-known that the Sato/Tanaka/Takeshita faction was the biggest faction overall and that it particularly colonized the Upper House. Our figures show that that faction’s median percentage of all Upper House prefectural winners (including the unaffiliated) was 29.6%. In the national district, they did even better, with a median of 36.4% of all winners belonging to their faction. The high water marks for the Sato/Tanaka/Takeshita faction were 41.2% of the prefectural winners (in 1992) and 63.2% of the national winners (in 1983). Overall, they had around 43% of the Upper House conservative seats in 1992.

These figures contrast sharply with the faction’s size in the Lower House. The median size for the Sato/Tanaka/Takeshita faction was 20.0% and there was much less variability in its size than in the Upper House, with the high water mark being only 28.2% (in 1992) and the low water mark only 13.8% (in 1962). Thus, as expected, the size of the largest faction in the Upper House was much less constrained than the size of the largest faction in the Lower House.

Progressive H1: Were Upper House LDP members with cabinet ambitions more likely to join factions than their less ambitious colleagues? Some evidence for this argument is provided by Sato and Matsuzaki, who argue that the Upper House factionalized as a consequence of what was going on in the Lower House, especially as regards the allocation of posts. By the early 1960s, there were three "keiretsu" in the Upper House: the "Seishin Club" (comprised of members loyal to Lower House factional leaders Sato, Kishi , and Kawashima), the "Mizuho Club" (with members loyal to Lower House factional leaders Kono, Fujiyama, Funada, and Murakami), and the Konwakai (with members supportive of Lower House factional leaders Ikeda/Maeo, Miki, and Ishii). According to Sato and Matsuzaki, these three "supra-factional" groups formed to compete for the three cabinet posts that were reserved for Upper House members. Thus, progressive ambition appears to have been at the root of Upper House factionalization.

Table 3 presents data that further support the hypothesis that Upper House LDP members with cabinet ambitions were more likely to join factions. Of those (66) who served in the cabinet at some point during their Upper House stints, 6.1% never affiliated with a faction. In contrast, of those (254) who never served in the cabinet during their Upper House careers, 16.5% never affiliated with a faction. This relationship is statistically significant at the .05 level by a c ² test.

Progressive H2: Another hypothesis regarding progressive ambition suggests that Upper House conservatives with Lower House ambitions might be more likely to sign up with factions than their colleagues without such ambitions, although we cast doubt on this hypothesis in our discussion above. Upon examining the data, it turns out that we cannot reject the null hypothesis that the affiliation rates were the same among these two groups. Of 414 Upper House conservative winners between 1962 and 1992, 63 (15.3%) later ran for the Lower House, while 351 did not. Of those who did eventually run for the Lower House, 15.9% never affiliated with a faction in the Upper House. Of those who did not later run for the Lower House, 12.0% never affiliated. The difference is not statistically significant in either direction. Progressive ambition of this form has no clear theoretical impact on factionalization (as noted above), nor is there any apparent empirical distinction.

Target H1: Evidence that the Upper House was colonized by factional bosses seeking extra votes in the LDP presidential election is available in narrative histories of the Upper House. Iseri claims that Sato Eisaku initiated the true factionalization of the House of Councillors in 1964, during his hard-fought battle with incumbent Ikeda Hayato for the party presidency. Sato sought to counter Ikeda’s Lower House supporters by recruiting Upper House members (see, e.g., Asahi Shimbun July 3,1964; Iseri 1988:44). Thus, by the late 1960s the early multi-factional alliances in the Upper House had broken up into groups with a one-to-one correspondence with Lower House factions (Sato and Matsuzaki 1986: 242; Bungei shunju 1968 [11]: 94-95). Tanaka Kakuei pushed the process of colonization even further in the early 1970s, in his efforts to usurp control of Sato’s faction, and later, in his battle with arch-rival Fukuda Takeo for party supremacy.

If Upper House factions were formed as a consequence of Lower House factional leaders’ presidential ambitions, then one should find smaller factions, those without a realistic shot at the presidency, less willing or able to bid high for factional recruits in the Upper House. And indeed, we find (see Table 4) that the smaller factions died out in the Upper House more quickly than they did in the Lower House, although there seems to have been a modest resurgence of small factions (mostly spinoffs from larger factions) in the Upper House starting in 1989.

Static H1: In order to investigate whether those prefectural MHCs who did face intraparty competition were more likely to affiliate with a faction than those who did not, we present a multivariate probit analysis that also bears on Contrast H1 and Progressive H1. The dependent variable in the analysis is a dummy variable equal to 1 if a given Upper House conservative in a given term was affiliated with a faction, equal to 0 otherwise. The independent variables consist of four dummy variables, one of which (CABINET) identifies MHCs who served in the cabinet during the term in question, and three of which identify different electoral incentives: NATIONAL_SNTV equals 1 for MHCs elected from the national district under SNTV (1965-1980); NATIONAL_LIST equals 1 for MHCs elected from the national district under list PR rules (1983-1992); and PREFECTURAL_INTRA equals 1 for MHCs elected from one of the prefectural districts who did face intra-party competition.

The results of the analysis are displayed in Table 5. Each of the coefficients can be interpreted as the difference in factionalization rate or probability between the baseline group of conservatives -- prefectural MHCs elected in districts without another conservative running -- and the various groups identified by the dummy variables. Our previous results hold up to multivariate scrutiny: MHCs who were in the cabinet or were elected under the list PR rules in force after 1983 were both more likely than the baseline group to join a faction; in contrast, national MHCs who were elected under SNTV rules were less likely to join factions than the baseline group. Each of these results is as expected theoretically and in accord with the bivariate evidence presented earlier.

As can also be seen, however, there is no tendency for prefectural MHCs who did face intra-party competition to join factions with greater probability than their colleagues who did not face such competition. Evidently, the occasional appearance of intra-party competition in the Upper House was not sufficient to produce the same sorts of adaptations as were observed in the Lower House, where intra-party competition could be expected year-in and year-out.

Incongruent bicameral systems—those using significantly different methods of election for their two legislative chambers—present a particularly attractive venue for studying the impact of electoral rules on party systems. Social cleavages and political culture—two of the most difficult factors to measure adequately in quantitative analyses—are held constant by studying differences between chambers in a single country at a single time (assuming concurrent or nearly concurrent elections). Thus, in examining such cases one can hope to get a cleaner view of how electoral incentives affect the formation of factions, parties and alliances (Cox 1997).

In this paper, we take advantage of the "natural experiment" provided by the very different electoral systems used for Japan’s Upper and Lower Houses. This "experiment" allows us to hold cultural factors, often credited as the underlying reason for Japanese factionalism, constant. The Upper House is just as Japanese as the Lower House, so a pure culturalist might say (with apologies to Gertrude Stein) that "a faction is a faction is a faction." On the other hand, an electoral incentives perspective predicts systematic differences in the nature of factionalism between the two houses.

Our analysis has canvassed a range of differences that one would expect between the two chambers, based on their differing electoral systems and differing positions in the political career structure of Japan. As predicted, we find that factionalization rates were lowest among national-district MHCs elected under SNTV rules, higher among prefectural MHCs, and highest among national-district MHCs elected from PR lists—with only the last group approximating the high rates found in the Lower House. We also found that MHCs with cabinet experience were more likely to have joined factions and that factional penetration of the Upper House was greatest among the bigger factions (those with a real chance at the party presidency), leading to the earlier demise of small factions in the Upper House. Finally, the staggered terms in the Upper House produced significantly more "same-faction-same-district" pairs than in the Lower House, with the result that the distribution of factional strengths in the Upper House was much more skewed than in the Lower House.

There are not many democracies with incongruent bicameralism, where the two chambers are elected under significantly different rules. Moreover, some of the incongruent cases are so classified due to differences in the size and apportionment of the districts used in the two houses, rather than to differences in the method of election. Incongruence due to size differences or malapportionment, although politically relevant, tends not to affect the party system as much as would incongruence in the rules of election proper. Focusing just on the cases with rule-based incongruence, opportunities to pursue studies similar to that undertaken here (albeit not necessarily focusing on factionalism) might arise in Australia, Bolivia, Brazil, Colombia, Poland, Spain, Switzerland and Venezuela. These cases are covered in a broad brushstroke fashion by Cox (1997) but only a few have more detailed studies (see, e.g., Jaensch 1983 on Australia; Shugart 1985 on Venezuela).

In addition to the opportunity to pursue more detailed electoral studies that exploit bicameral incongruence, we should also note that a similar research strategy is available in exploiting incongruence between local/regional and national elections. There are a variety of studies emerging that demonstrate the importance of local/regional factors on national-level outcomes (e.g., Jones 1997; Samuels 1998) but there has been little on how local/regional systems "fit" with national systems. In Chile, for example, municipal elections are held under a high-magnitude PR system while national elections are held in two-seat districts. Part of the reason for the persistence of Chilean multipartism in the face of a system engineered to produce bipartism may be that the "feeder system" into national politics remains thoroughly multiparty.