¹ Tsebelis, George, Nested Games: Rational Choice in Comparative Politics (Berkeley: University of California Press, 1990), chap. 5.Google Scholar

² Kenny, Patrick and Rice, Tom W., ‘The Relationship Between Divisive Primaries and General Election Outcomes,’ American Journal of Political Science, 31 (1987), 31–44.CrossRefGoogle Scholar

³ The original Battle of the Sexes involved a woman wishing to go the ballet, a man wishing to go to a boxing match, both preferring the other's company at the less-preferred entertainment to going solo to the more-preferred entertainment.

⁴ See Raymond Christensen, ‘Strategic Imperatives of Japan's SNTV’ (unpublished paper, University of Kansas, 1994), for a discussion of some similar problems besetting the efforts of the Japanese opposition parties to co-operate electorally.

⁵ Gallagher, Michael and Marsh, Michael, eds, Candidate Selection in Comparative Perspective (London: Sage Publications, 1988).Google Scholar

⁶ The Japanese did impose an additional requirement: in order to win a seat, a candidate had to garner more than a legally defined ‘minimum vote’. The minimum was set at such a low value, however, that no candidate who finished in the top M places in a district failed to attain it. District magnitudes in Japan formerly ran from one to six, inclusive, with most of the districts being in the range from three to five.

⁷ Insisting on continued support was valuable not only in bluffing factional opponents but also in recruiting new members. A faction that did not continue supporting its non-incumbents after they had lost out in the quest for the party nomination would find it more difficult to recruit good candidates at the next election.

⁸ Cox, Gary and Niou, Emerson, ‘Seat Bonuses Under the Single Non-Transferable Vote Systems: Evidence from Japan and Taiwan’, Comparative Politics, 26 (1994), 221–36.CrossRefGoogle Scholar

⁹ In practice, the Electoral Strategy Committee's decisions are reviewed by the party's Executive Committee and sometimes by the senior officials (San Yaku) of the party. See Shiratori, Rei, ‘Japan: Localism, Factionalism and Personalism’, in Gallagher and Marsh, eds, Candidate Selection in Comparative Perspective, 169–89, at p. 172.Google Scholar

¹⁰ The 1955 party constitution had specified the composition of the Electoral Strategy Committee in considerable detail. At the Party Congress in 1960, the party amended the constitution to require as members only the party president, vice president, and secretary general; nine others were to be appointed by the president, for a total of twelve members (see Liberal Democratic Party, Juyu minshuto nijunen no ayumi [The LDP's Twenty Years] (Tokyo: Liberal Democratic Party Press, 1987), p. 49.Google Scholar This section is based on a previous working paper: see Cox, Gary and Rosenbluth, Frances, ‘Factional Competition for the Party Endorsement: Japan's Liberal Democratic Party’ (unpublished paper, University of California, San Diego, 1994).Google Scholar

¹¹ Liberal Democratic Party, Juyu minshuto junen no ayumi [The LDP's Ten Years], (Tokyo: Liberal Democratic Party Press, 1966)Google Scholar; Liberal Democratic Party, Juyu minshuto nijunen no ayumi [The LDP's Twenty Years] (Tokyo: Liberal Democratic Party Press, 1975)Google Scholar; Liberal Democratic Party, Jiyu minshuto toshi, shiryo hen [A History of the Liberal Democratic Party], Vol. 3 (Tokyo: Liberal Democratic Party Press, 1987).Google Scholar

¹² Liberal Democratic Party, The LDP's Ten Years, p. 500.

¹³ Although excluding those indicted may have been easier to agree on than other exclusionary principles, it was not easy on an absolute scale, largely because candidates were worried of running foul of the very strict Japanese campaign regulations (see Thayer, Nathaniel, How the Conservatives Ruled Japan (New York: Columbia University Press, 1969), pp. 122–5).Google Scholar The promise to consider the recommendations of the local branches is the only endorsement principle that does not bear directly on the LDP's goal of limiting the number of candidates. For a discussion of the logic behind this principle, see Thayer, , How the Conservatives Ruled Japan, pp. 121–2.Google Scholar

¹⁴ ‘Shinogi kezuru “habatsu senkyo”’ [The cutthroat competition of the ‘factionalized election’], Asahi shimbun, 14 12 1969.Google Scholar

¹⁵ ‘Ranpatsu sareta toseki shomei’ [Party IDs are over-issued], Asahi shimbun, 12 09 1979.Google Scholar

¹⁶ ‘Hiki orosih kosaku ni hanpatsu’ [Resisting forced withdrawal of support], Asahi shimbun, 17 01 1967.Google Scholar

¹⁷ Sato, Seizaburo and Matsuzaki, Tetsuhisa, Jiminto seiken [The LDP Administration] (Tokyo: Chuo koron sha, 1986).Google Scholar

¹⁸ Fukui, Haruhiro, ‘The Liberal Democratic Party Revisited: Continuity and Change in the Party's Structure and Performance’, Journal of Japanese Studies, 10 (1984), 385–435.CrossRefGoogle Scholar

¹⁹ While this conventional wisdom does manifest itself in print from time to time, we are aware of it primarily through conversations with various Asahi Shimbun journalists.

²⁰ The data used in this and all other analyses conducted in this Note are from Reed, Steven's compendium Japanese Election Data: The House of Representatives 1947–1990 (Ann Arbor, Michigan: Center for Japanese Studies, 1992).CrossRefGoogle Scholar We restrict attention to the top five factions – those founded by Sato, Ikeda, Kishi, Kono and Miki – because they are by far the most important and because identification of the non-incumbents that they endorsed is probably more complete. The results do not, in any event, change materially if non-incumbents backed by other factions are included in the analyses.

²¹ Curtis, Gerald, Election Campaigning, Japanese Style (New York: Columbia University Press, 1971).Google Scholar

²² Probit coefficients are of course not directly interpretable in the way that ordinary regression coefficients are, but standard techniques can be employed to translate the coefficients into probability impacts. The impacts we report are for a hypothetical candidate whose initial probability of endorsement equals the 1960–90 endorsement rate of 0.74, and who then faces a change in mainstream status, a change in the number of incumbents in his or her district, and so forth.

²³ INCSPER measures the number of non-incumbent endorsed candidates per seat. This choice of unit means that, if one wishes to discover how much one more incumbent hurt the typical non-incumbent seeking endorsement, then one must first divide the coefficient estimate by the magnitude of the district in which the candidate ran. This division is necessary because ‘one more incumbent’ means one third more incumbents per seat in a three-seat district, a quarter more incumbents per seat in a four-seat district, and so on.

²⁴ It should be noted that ‘one more mainstream incumbent’ counts once in INCSPER and once in MAINSPER; SO the total effect is found by adding the coefficients of INCSPER and MAINSPER, then dividing by district magnitude. The additional impact of a mainstream incumbent is thus 0.08 in a three-seat district, 0.06 in a four-seat district, and 0.04 in a five-seat district.

²⁵ The only coefficient that does change significantly is that attached to the mainstream status variable: it declines from 0.30 to 0.21. If one includes both the PM dummy variable and the SECGEN dummy variable, the results change hardly at all, with the PM dummy's coefficient small and insignificant, the SECGEN variable's coefficient and standard error about the same size as in the last model in Table 1.

²⁶ Note that, since the Takeshita faction held the secretary generalship from 1983 to 1990, the TAKESHITA dummy variable is almost equal to the SECGEN dummy variable in the 1980–90 period. So it is not surprising that re-running the 1980–90 regression with SECGEN rather than TAKESHITA yields virtually identical results.

²⁷ Cox, Gary and Rosenbluth, Frances, ‘Anatomy of a Split: The Liberal Democrats of Japan’, Electoral Studies, 14 (1995), 355–76.CrossRefGoogle Scholar