¹ Interestingly, the rate of decline in Liberal Democratic Party support has itself decreased in recent elections. Voting patterns in House of Councillors elections have generally been comparable to those reported here for House of Representatives contests. The decline in Liberal Democratic support in 1958–72 can also be seen as an extension of diminutions in the conservative vote from the 1953 election on. Still, multiple changes in party labels in the conservative camp before 1955 make precise analysis of the individual voting behavior associated with this general trend highly problematic.

² Indeed, in recent elections the Liberal Democrats came close to losing their control of the Diet. In the 1972 House of Representatives election, the LDP received only a 26-seat majority in the 491-member body. However, their ranks were later increased by affiliation of several successful independent candidates. (See Senkyobu, Jichisho, Shūgiin Giin Sōsenkyo, Saikō Saibansho Kokumin Shinsa Kekka Shirable [Tōkyō, 1972], p. 14.Google Scholar) In the 1974 House of Councillors election, the story was even more grim for the conservatives; the LDP won only one-half the seats. Once again, a handful of independent candidates came to the rescue. (See Shigeki, Nishihira, “Dai Jukai Sangiin senkyo no sūryō bunseki,” Seiji Kōhō, Feb—Mar 1975, pp. 16–31Google Scholar.) Although typically there are some conservative candidates who run as independents for various reasons but can be counted on to return to the party after the election, the closeness of the LDP to defeat—especially in the House of Councillors contest—is striking.

³ Unfortunately, this basically psychological model of political behavior has thus far been applied only to attitudes toward political objects— e.g., issues, candidates, parties. There is an important gap in the evidence available on Japanese electoral behavior; the attitudes of Japanese toward social objects—e.g., their community, face-to-face groups and personal relationships, and the political content of communications with other persons in these categories—are not measured, even though most students of Japanese politics and society believe they are important to political behavior.

⁴ The years 1958–72 were chosen in order to examine party support patterns after the main conservative and progressive parties were merged in 1955. In some of the later analysis—specifically that of individual constituencies—the period 1960–72 was used, since that period seemed to bracket the major periods of change up until now at that level.

⁵ As will become abundantly clear, other things are not always equal. In addition to fluidity at the constituency and individual level being masked by the national figures, population changes can be important factors in electoral behavior. As some cohorts depart from the electorate and others enter, major shifts in voting support can occur. Also population movements occur between constituencies, as we shall note.

⁶ However, recent research on European party support patterns (using slightly different statistical measures) indicates that they are slightly more fluid in the Japanese than in the European case. As far as these national aggregate patterns go, Japan may not be far from the norm for industrialized societies. (Also, recent fluidity in voting patterns in countries like the Netherlands and Denmark raises important questions about the inherent stability of European voting behavior.)

⁷ In using the longer time period for these calculations, I sought to best capture long-term partisan change trends independent from the effects of individual candidacies, which can be conspicuously present at the constituency level. The calculations were similar to those applied to the national interelection differences: the difference between specific parties' shares of the vote in 1960 and 1972 was calculated for all parties, then halved; and the percentage change in the independent vote was subtracted from the resulting figure. It should be remembered that this index refers to movements in the aggregate vote, not to the behavior of individuals.

⁸ Within this stable group, 22 districts actually demonstrated change indices of 5% or less. This is a remarkably low figure, given the natural fluidity resulting from changes in the composition of the electorate and changes in candidates in some places.

⁹ For the identity of the specific constituencies in each category, see Table III.

¹⁰ The 1967 Ward and Kubota General Election Study data set was used in the analysis. The 1963–67 interval tended slightly toward the high side with regard to change, according to the figures in Table II on national aggregate change patterns. This reflected, among other things, the effects of the entry into Lower House elections of Clean Government Party candidates.”

¹¹ Persons who did not have the franchise in 1963, those who abstained in both elections, and persons who could not answer the questions on party voting choices were excluded from the analysis.

¹² Actually, the element of instability introduced in the aggregate figures by the physical change of the electorate, which is not reflected in these individual figures, makes the contrast even more dramatic. Two more things need to be said about this information on individual change: Persons who voted for different parties in 1963 and 1967 did not necessarily adopt a new party loyalty in so doing. Rather, we have here evidence of mainly short-term change, which may or may not lead to long-term change. This makes it difficult to relate the information on individual behavior to that on constituency patterns, which sought to identify long-term shifts in preferences. Also, although I have indicated that abstentions introduce an important element of potential fluidity into voting choices, we have no idea at present how these choices are aggregated into electoral decisions at any particular time. Abstentions obviously become an actual source of electoral change only when the former abstainers or contemporary abstainers add or subtract from the support of particular parties in such a way as to make an important impact on election results.

¹³ In addition to noting the different fluidity rates in different constituencies (Table III), I examined the individual reports of party preferences in 1963 and 1967 in constituencies having different levels of change. Unfortunately, individual change in 1963–67 did not reflect as dramatically as might be expected the overall changes in voting behavior in different kinds of constituencies. For this reason, “estimates” of individuals' propensity to change in the districts where the greatest change occurred were not too revealing.

¹⁴ Not very many Japanese reply directly that they consider national issues in their voting choices; see my The Political Culture of Japan (Berkeley: Univ. of California Press, 1974), pp. 108–12. Still, Japanese are knowledgeable about some issues, and some recent research indicates that these do play a role in voting decisions in that there are positive and significant correlations between issue positions and party support; see Shinsaku, Kōhei, “Political Behavior of the Japanese People,” NHK Public Opinion Research Institute, 1972Google Scholar, and “Tōhyō kōdō no shakai shinrigakuteki kenkyū,” Nihon Shakai Shinri Gakkai (Dai 11 kai taikai) paper.

¹⁵ The sum of all candidates in the 1958, 1960, 1963, 1967, 1969, and 1972 general elections was 5,500, according to calculations using data from Jichisho (n. 2 above), p. 5. If multiple candidacies by the same individuals are eliminated, the figure is reduced to 2,933.

¹⁶ There is some secondary literature on these and related topics in Japanese. See Shigeki, Nishihira, Nihon no senkyo (Tōkyō: Shiseidō, 1972), especially pp. 16–28Google Scholar, as well as the separate chapters on specific elections; and the NHK Yoron Chōsajo report “Sōsenkyo no bunseki (Tōkyō, 1972).

¹⁷ Obviously, we are interested here more in long-term changes in partisan attachments.

¹⁸ Whenever contemporary Japanese are asked to rank issues with regard to their perceived im-portance, domestic economic issues outweigh international problems and foreign policy concerns. In some instances, local issues outweigh national problems in importance.

¹⁹ In a secondary analysis of the Ward and Kubota 1967 data set, I found that persons who felt their economic situation had improved had turned to Liberal Democrats from other parties in disproportionate ratios.

²⁰ The information presented in Table IV did fluctuate monthly, with the result that there is a certain unevenness in trends, such as indicated in the “policies and capability” column for 1968 and 1970. But the overall trends were quite consistent, and supportive of the comments in the text.

²¹ The actual numbers of Clean Government and Communist Party supporters were still quite small relative to those for the more established parties, in part simply because the Clean Government and Communist parties still attract fairly new supporters proportionately. But the phenomenon of “concealed partisanship,” i.e., the tendency to avoid mentions of loyalties to parties that are not generally socially approved, may have been present, too. For citations of this phenomenon in France, see Philip Converse and Georges Dupeux, “Politicization of the Electorate in France and the United States” in Campbell, A. et al. , Elections and the Political Order (New York: John Wiley, 1966), pp. 169–91Google Scholar.

The NHK research group has reported findings somewhat analogous to those I have discussed here. Thus, Clean Government and Communist Party supporters felt that the government was responsible for price rises far more often than did followers of other parties; see “Sōsenkyo no Bun-seki” (n. 16 above), pp. 160–63. As can be seen from the figures in Table V, there is a slight downturn in the net satisfaction index in 1971–72 relative to earlier years. In 1973–74 there was a dramatic increase in net dissatisfaction, which—if our analysis is correct—should lead to even more movement to the Japan Communists and the Clean Government parties in coming elections.

²² Some of the distortion present in Japan's contemporary electoral system can be seen from the difference between the votes received by successful candidates in urban and rural districts. For example, in 1972, winning candidates in Tōkyō's most populous district (Tōkyō 7) received 148,000–208,000 votes, in contrast with those from Ehime 3, who gained 47,000–50,000 votes. These two districts represent extreme cases, but urban populations are generally underrepresented relative to the rural sector.

²³ Information taken from Terry MacDougall's “The Progressive Alternative in Japanese Local Political Leadership” (paper presented at 1973 annual meeting of the Association for Asian Studies), esp. p. 26 and Appendix 1, and based on the party ties of identifiable members of the progressive mayors' association. It is difficult to carry the research further, since many progressive mayoral and gubernatorial candidates ran as independents. For this reason, they cannot be distinguished from conservative independents in election reports (which could otherwise be used to better establish the progressive local executive—national voting pattern linkage).

²⁴ In fact, a clearer link could be drawn between progressive local politics and national voting patterns in the suburbs. Cities in the suburbs had mayors, whereas the urban portions of Tōkyō did not have mayoral races.

²⁵ The cited progressive mayorships do not include all cities in the relevant House of Representatives constituencies—in some cases, not even a majority. Thus the inference about the link between the presence of local progressive executive movements and shifts in the national vote is tenuous in some cases. This is an area where more research is needed.

²⁶ For a brief discussion of these movements, see Simcock, Bradford, “Environmental Pollution and Citizens' Movements” in Area Development in Japan (Tōkyō, 1972), pp. 13–22Google Scholar. The Mishima campaign is also discussed briefly in Jack Lewis, “Coalition Formation in a Local Multi-party System: Implications for Political Change in Japan,” paper presented at the AAS 1973 annual meeting.

²⁷ Some studies of the relationship between citizens' movements and political parties do not support the linkage I have asserted here. Neither have recent studies of this subject by Margaret McKean and Brad Simcock, which will be published in the forthcoming study of Japanese local politics edited by Kurt Steiner, Scott Flanagan, and Ellis Kraus. Still, the relationship between the presence of citizens' movements and marked shifts in voting patterns in areas where no other plausible explanation for sudden change exists is striking. (For the work of McKean and Simcock, see respectively “Political Socialization through Citizens' Movements” and “Citizens' Movements and Opposition Politics in Japan,” papers presented at SSRC conference on Local Opposition in Japan, June 1976 at Wrightsville Beach, N.C.)

²⁸ Although many young people in Japan, as elsewhere, presumably inherit the party loyalties of their parents, there are some unknown limits to the effects that physical change in the population may have. Recent research on the British electorate does indicate that physical replacement of the electorate can play an important role in partisan change, and both processes of inter-generational transmission of partisanship and discontinuities between generations need be given their fair due in longitudinal analyses of electoral behavior. For comments on the inter-generational transfer of party loyalties in Japan, see Kubota, Akira and Ward, Robert E., Comparative Political Studies, 3 (1970), pp. 140–75CrossRefGoogle Scholar; for analysis of the effects of physical replacement, see Butler, David and Stokes, Donald E., Political Change in Britain (New York: St. Martin's, 1969), pp. 263–74Google Scholar.

²⁹ There was also in the “new” cohort a substantial increase in the ratios of young persons who supported no party, in contrast with persons in this age group a decade ago—an additional factor that may contribute to electoral fluidity in Japan. In this, Japanese youths are similar to their counter-parts in other industrialized societies.

³⁰ The lack of a decline in LDP support shown among cohorts over 30 years in age in 1971 (see Table VI) does not necessarily contradict the trends shown by information on voting in elections, since former LDP voters who now support other parties could actually be concealed within the no-party-support category. (Obviously, a substantial increment of voting change between 1960 and 1972 could have resulted from the lower LDP support rates among young persons.)

³¹ There is a paucity of evidence on these important longitudinal dimensions of political behavior. For some societies—notably Scandinavia, the Low Countries, Austria, and Italy—there is ample confirmation of the relationship between party identification and the vote, and linkage between social group memberships and party choices. And self-reported consistency in voting habits is extremely high among partisans in several European countries. But only fragmentary directevidence is available on the continuity of party loyalties across time, and in many cases the time periods are not long enough to actually confirm the hypothesis that long-term psychological loyalties to political groups actually exist. Still, from the extreme stability shown by aggregate party support patterns in many countries, they are typically inferred to be present.

For studies of the vote within social groups in Europe and elsewhere: Rose, Richard, Comparative Electoral Behavior (New York: Free Press, 1974)Google Scholar. For information on the high consistency of partisan choices over time based on recall: inter alia, Angus Campbell and Henry Valen, “Party Identification in Norway and the United States” in Elections (n. 21 above), pp. 245–68. For evidence on the longitudinal durability of party identification: Political Change (n. 28 above), pp. 293–303; and Borre, Ole and Katz, Daniel, “Party Identification and Its Motivational Base in a Multiparty System: A Study of the Danish General Election of 1971,” Scandinavian Political Studies, 8 (1973), p. 78CrossRefGoogle Scholar.

³² Some research on both American and British electorates' party loyalties postulates that partisanship is learned early in life, and increases in intensity over the life span. One reason given for the deepening intensity of attachments among older people is the effect of consistent voting experiences on attitudes. Somewhat similar learning and conditioning patterns could be anticipated for candidate loyalties in Japan, except for the important fact that candidates are sometimes less durable features of the electoral environment than parties.

³³ See Howell, Susan, “System Effects on Legislator Visibility in Five Democratic Countries,” Comparative Political Studies, 8 (1976), pp. 413–36CrossRefGoogle Scholar.

³⁴ The importance and longevity of candidates and parties as electoral objects still varies among parties and over time. In the 1950s, there were many conservative candidates who had outlasted their parties; but the proportions are reversing somewhat now that the Liberal Democratic Party has been around for two decades. The relative relationship is similar in the case of new parties like the Democratic Socialists in 1959 and the Clean Government Party in the 1960s. In contrast, the Socialist and Communist Parties have been more stable political objects than many of their candidates.

³⁵ For the wording of the relevant questions, see the 1967 Ward and Kubota survey protocol or the ICPSR-distributed codebook for the survey data set.

³⁶ Analysis employing the alternative measure of candidate ties—reports that respondents would never consider another candidate—showed the same kinds of patterns. Stable voters were attached to candidates more than changers, but less than half of the stable voters indicated such an attachment. It is possible that the measures employed here did not completely tap the attitude I sought to identify, i.e., habitual ties to a particular candidate. Also, it is plausible that candidate turnover may have made some people less decisive and committed than would have otherwise been the case. But, using the information at hand, there was no way to eliminate persons who were faced by ambiguous alternatives due to the disappearance of a favored candidate.

³⁷ Reporting that one would never vote for another party also was associated more with stable voting patterns, although at somewhat lower levels than was the case of persons who simply indicated that they “liked” a particular party.

³⁸ See my “Party Loyalties and Party Saliency in Japan,” Comparative Political Studies, 8 (1975), pp. 32–57CrossRefGoogle Scholar; also Ichiro, Miyake, Tomio, Kinoshita, and Toshiichi, Aiba, Tōbyō kōdō no kenkyū (Tōkyō: Sōbunsho, 1967), p. 73Google Scholar.

³⁹ There was the possibility that some people might be tied to parties, others to candidates, and others to social classes, so that the cumulative effects of political loyalties might be greater than Table VII suggests. However, analysis showed that this was not the case: political ties were overlapping and reinforcing, rather than cumulative.