Still valuable? Reconsidering the role of authoritarian values among Japanese voters

Abstract The Liberal Democratic Party (LDP) has dominated Japanese politics since its foundation in 1955. The party's ability to support the interests of small farmers and its commitment to the middle-class through redistributive economic policies have been regarded as among the primary factors leading to its prolonged hegemony. Yet, the LDP has occasionally relied on non-economic political appeals addressing rather conservative issues such as Japan's military force and the country's traditional values. These appeals have significantly intensified under the premiership of Junichirō Koizumi and his successor Shinzō Abe, whose authoritative leadership styles and nationalist agendas reveal the relevance of non-economic values. This study focuses on the role of authoritarian values in shaping vote choice in twenty-first century Japan. Previously emphasized in The Japanese Voter, the authority–liberty dimension has gradually lost prominence in recent models of voting behavior, where socio-economic factors have had a central role in explaining vote choice. Our inquiry places new emphasis on this value dimension and provides evidence for the enduring relevance of individual authoritarian dispositions in models of vote choice.


Introduction
Empirical research on voting behavior constitutes one of the most developed political science fields in Japan. As we shall see below, much emphasis has been placed on the role of economic evaluations in explaining vote choice in recent decades. This perspective partly derives from the burgeoning of economic voting models developed in the rational choice framework. Yet, despite the general understanding of the importance of security issues and conservative agendas in structuring ideological competition in contemporary Japan, aspects such as non-economic values have been neglected in recent studies of voting behavior. We aim to fill this gap in the literature and rectify discrepancies between the general reading of contemporary Japanese politics and the academic literature.
The emphasis on socio-economic explanations stems from the characteristics of Japan's party system. Japan is one of the most prominent cases of dominant party systems among the established democracies. The right-wing Liberal Democratic Party (LDP) has been in power almost uninterruptedly since 1955, holding the largest share of seats in the House of Representatives with the only exception of the 2009 election. 1 The stable linkage between party system configurations and the cleavages present in society as conceived in classic cleavage theory (see Lipset and Rokkan, 1967) suggests that party system change may result from changes in cleavage structures. Indeed, the evolution of Japan's party system seems to reflect the social transformations that have taken place since 1955, namely, changes in the occupational structure amid rapid economic growth reduced perceptions of class conflict, thereby accelerating the decline of the Japan Socialist Party (JSP), and the generational turnover weakened value oppositions between the pre-and post-war generations (Schmidt, 2003). Furthermore, the LDP was able to take on the role of social peacemaker by committing to the 'middle-class model,' wherein the fair redistribution of national wealth and the promotion of social mobility constituted the pillars of the LDP's economic policies (Chiavacci, 2006). Not only did the success of the LDP's middle-class model inhibit the potential of class conflict, but the party also managed to strategically avoid divisive issues such as constitutional revision and rearmament, focusing instead on economic growth and redistribution (Reed, 2005). Yet, since the charismatic Junichirō Koizumi took the party reins in 2001, the LDP has experienced a drift to the right which the literature characterizes in various ways, for instance in terms of 'religious nationalism' (Mullins, 2012) or 'illiberal' and 'authoritarian' turn (Nakano, 2016). Indeed, under Koizumi and Abe, the LDP's plan to revise Japan's pacifist constitution and the promotion of historical revisionism and patriotic education regained prominence in both the political discourse and the policy agenda. 2 We argue that the revival of nationalist discourses echoing the esthetic ideas of collective patriotism and shared identity as well as the authoritative leadership styles of Koizumi and Abe are indicative of the relevance of non-economic values in shaping voter alignments. In hindsight, the 'drift to the right' in the recent decades partly coincided with the burgeoning of economic voting theories and the flourishing of studies that emphasize the role of the economy rather than value competition (see Kobayashi, 1991;Hirano, 1993Hirano, , 1994Hirano, , 1998Tsutsumi, 1997;Endo, 2009;Taniguchi, 2016;Ohmura, 2018, among others). Thus, despite the general impression of a hawkish turn under Koizumi and Abe, these studies may have overlooked the relevance of non-economic values in contemporary Japanese politics. To this end, our inquiry reconsiders the role of authoritarian values, a key value dimension previously introduced and analyzed by Scott Flanagan in The Japanese Voter (1991). As detailed below, we propose an empirical reassessment of the relevance of authoritarian values based on the behavior of Japanese voters from 2003 to 2019.
The article proceeds as follows. Section 2 discusses the types of social cleavages and voter alignments examined by previous studies and summarizes the main models of vote choice in the literature on Japanese voting behavior. Section 3 develops the main argument of the article and presents the hypothesis. In Section 4, we illustrate the data and methods used to evaluate the role of authoritarian values in shaping vote choice. In Section 5, we present the empirical results and discuss their implications. The article concludes with a summary of the key findings and some considerations on the contribution of our study to the literature on voting behavior in Japan and comparative party politics in general.

Social class, values, and other factors in Japanese voting behavior
Before discussing in detail the role of authoritarian values in voting behavior, let us first review a brief chronology of the main academic contributions in the field. Early studies of voting behavior in post-war Japan evidenced the presence of clear patterns of class voting, with the independent occupational classes voting for LDP and the working classes predominantly casting their ballot for the JSP or the Japanese Communist Party (JCP). Miyake (1985) defined this political divide as the dichotomy between jimae (self-sufficiency) and hi-jimae (dependency), whereby independent farmers, entrepreneurs, and the self-employed belonging to the former group voted conservative, and the latter category comprising clerks, workers, and employees was generally closer to the left-wing parties.
In spite of the growing industrialization and massive inflows of new workers into urban areas, these patterns of party support based on social class rapidly declined starting in the 1960s, when the JSP began to lose votes from its traditional constituencies (Maruyama, 2007: 33). The decline of the JSP has been attributed to the increasing competitive pressures from minor parties under Japan's former multi-member electoral system (Kohno, 1997;Mori, 2001) and to the LDP's ability under Ikeda to make credible redistributive commitments while avoiding direct confrontation with the Socialists on divisive issues (Watanuki, 1991: 62-64). The limited explanatory power of social class (see Flanagan and Richardson 1977;Miyake, 1977;Watanuki, 1991) suggests that Western-style patterns of class voting are relatively less pronounced, if not absent, in Japan.
Alternative theories of voting stress the role of values as better suited for explaining vote choice. Value competition emerged in Japan in the wake of socio-political transformations brought about by democratization. The new 1947 Constitution drafted under the supervision of the US Occupation replaced the former imperial rule with a democratic system based on free elections, the separation of powers and the rule of law, turning Japan into a full-fledged liberal democracy.
The rapid transformation into a multiparty democracy under the US guidance contributed to polarizing the Japanese public around competing systems of values (Flanagan, 1991). Watanuki (1967) argued that high-pace economic growth alongside the general improvement of citizens' well-being further reduced the space for competition on economic issues in favor of political divisions based on values, a dimension of conflict the author defined as 'cultural politics. ' Flanagan (1991) examined the different subdimensions of value change and argued that Japan's democratic transition and the shift from an agrarian to industrial society favored the rise of a new value cleavage that he characterized as of authority vs liberty (A-L). In his analysis, Flanagan found substantial correlations between the A-L scale and partisan preferences in the post-war period, with LDP supporters having more pronounced authoritarian tendencies than supporters of the opposition parties. 3 Partisan conflicts on economic issues seemed to be overshadowed by value competition, as long as the Japanese economic miracle continued to deliver growth and prosperity to the general population. Yet, since the 1980s non-economic values have gradually lost prominence in models of Japanese voting behavior, as suggested by the growing number of studies on economic issues. For example, Reed and Brunk (1984) found that while economic factors had little effect on the vote in times of economic expansion, substantial relationships between economic variables and election results have become more evident since the 1975 recession. Wade and Owens (1992) showed that the patterns of economic voting in Japan in the 1980s resembled those of Western democracies, with significant effects of unemployment levels on the vote. In his comparative assessment of competing models of voting behavior, Nakamura (2003) observed that party-voter policy proximity and individual evaluations of subjective economic well-being (pocketbook voting) were better suited to explain vote choice in the 1983 Upper House election. Similar findings have been reported by a number of studies, that emphasized value-free factors such as evaluations of governments' past performance, subjective wellbeing, and future expectations as powerful predictors of vote in the elections held during the 1980s and 1990s (see Hirano, 1998: 33). Lastly, Ohmura (2018) provided further evidence that individual evaluations of the economy have become increasingly important drivers of voting behavior.
Indeed, more recent studies suggest that substantial changes in the Japanese party system such as the 1994 electoral reform, the decline of the LDP, and the rise of the Democratic Party of Japan (DPJ) did not reflect the emergence or reactivation of deep socio-cultural cleavages in Japanese society. Miwa's (2018) assessment of the value preferences of Japanese voters and party representatives found only slight differences across the main party electorates. Chiavacci (2010) examined the electoral success of the DPJ in 2009 and argued that the party was able to present itself as the new champion of 'shared growth,' campaigning on the same issues and for the same votes from rural areas that once secured landslide victories to the LDP. This suggests much greater salience of redistributive issues, with little room for partisan conflicts on ideology and values. The idea of a general tendency toward the de-ideologization of Japanese politics finds further support in Takenaka's (2014) analysis of the declining role of the conservative vs progressive ideological dimension and its weakening association with party preferences across generational cohorts.
Likewise, recent accounts of the success of the LDP, which has steadily returned to power since 2012, seem to take a rather economic perspective and to attach less importance to value oppositions: Umeda (2019) attributes the LDP's hegemony to its adaptability to new environments, whereby among others, the party's commitment to protecting the interests of small farmers and businesses from globalization was a crucial factor for power consolidation. A similar reasoning was proposed by Lind (2018) in her discussion of Japan's seeming immunity to the rise of populism. Lind argued that the preconditions for the success of populism were missing in Japan because the LDP successfully resisted pressures toward liberalization by raising non-economic trade barriers and limiting immigration, while protecting Japan's vulnerable sectors from global competition.

Bringing values back
Although economic considerations are certainly indispensable to explain the patterns of voting, we argue that recent studies emphasizing economic and redistributive issues may have overlooked the role of values. Our inquiry reassesses the relationship between value orientations and partisan support in contemporary Japan and reconsiders the place of non-economic values in the literature of Japanese voting behavior. We do not reject the economic arguments reviewed above but argue that the role of values has not been assessed adequately in addition to the mainstream accounts based on economic evaluations and policy proximity. Given the changing environment of party competition on the supply-side and the changing nature of perceived ideology (Jou and Endo, 2016) on the demand-side in recent decades, it seems pertinent to assess the role of values and economy together in a systematic manner.
In the last 40 years, value oppositions have indeed gained prominence in the analysis of political behavior: the unfolding of Inglehart's Silent Revolution since the 1980s has profoundly transformed the pre-existing patterns of political divisions in the developed democracies, raising the salience of a New Politics dimension in addition to, or in replacement of, the conventional left-right continuum (see Kitschelt and Hellemans, 1990). Scholars have argued that the rise of left-libertarian values such as tolerance toward minorities, self-reliance, and individual freedom has triggered a counter-revolution from the right side of the political spectrum, whereby radical right parties emphasizing the values of authority and hierarchy have capitalized on concerns over immigration, social order, and security (Ignazi, 1992;Hino, 2003). Likewise, the Cultural Backlash theory characterizes the current surge of authoritarian populism as a response from the older and traditionalist strata of society who feel that their values are being threatened by cultural shifts toward post-materialism (Norris and Inglehart, 2019). These radical right parties typically embrace nativist and authoritarian views, in open defiance of liberal democracy (Mudde, 2007), and cater to older and less-educated voters expressing concerns about immigration, identity, security, and adhering to authoritarian values.
In the West, recent political developments such as the victory of Donald Trump in the 2016 US election, the Brexit referendum, and the emergence of successful right-wing populist leaders, for instance Marine Le Pen in France and Matteo Salvini in Italy, are indicative of a surge of authoritarian populism and growing attractiveness of far right discourses. In contrast, Japan has not experienced disruptive changes in the patterns of party competition, as the prolonged and seemingly unchallengeable hegemony of the LDP would suggest. However, some scholars have argued that nationalist rhetoric has been on the rise within Japan's conservative elites (e.g., Matthews, 2003;Sasada, 2006Sasada, , 2010Mullins, 2012).
Since the premiership of Junichirō Koizumi (2001Koizumi ( -2006, recent political trends in Japan suggest the potential for the reemergence of value competition among the Japanese public. Under Koizumi and Abe, Japan experienced a right-wing or neo-conservative turn (see Takahashi, 2010;Nakano, 2016;Ramirez, 2018) that revitalized previously quiescent issues such as national identity and pride and Japanese traditional values. Arguably, economic reforms are among the key policy innovations for which Koizumi and Abe have been known internationally: Koizumi's plan to privatize the Japan Post and the Abenomics reforms were central issues in the electoral campaigns of 2005 and 2012, respectively. However, Nagy (2014) argued that Koizumi and Abe utilized nationalist rhetoric as a strategic leverage to consolidate their leadership position within the LDP and to secure public support for their economic reforms. Besides the increasing attention devoted to national security among the LDP candidates (Catalinac, 2016), these nationalist appeals have primarily regarded divisive topics such as historical revisionism, patriotic education, and rearmament. In addition, the controversial issue of prime ministers' visits to the Yasukuni Shrine, which were resumed under Koizumi after a substantial interruption since 1985 (Deans, 2007), carries a highly symbolic meaning with regard to Japan's role in World War II.
Indeed, the ideological proximity of Abe's LDP with the widespread phenomenon of online right-wing activism (netto uyoku) (Schäfer et al., 2017) and the outstanding number of LDP representatives, including Abe himself, who are among a group of members of parliament who support the ultra-conservative lobbying group Nippon Kaigi (the 'Japan Conference') are revealing of a changing political climate. Higuchi and Matsutani (2013) examined the radical right phenomenon from the demand-side and found that nationalist and authoritarian tendencies are common among supporters of both Abe and the ultra-conservative Ishihara Shintarō, with the latter also attracting voters with xenophobic attitudes, similarly to the European radical right. A few books (see Tsukada, 2017;Tanabe, 2019;Oguma and Higuchi, 2020) have been recently published with titles including the oft-used expression 'drift to the right' (ukeika), examining to what extent and why this rightward drift is taking place in Japan. Among these, Tanabe (2019) and Matsutani (2020) relied on various survey data to demonstrate that nationalism and xenophobia against China and Korea are generally on the rise in Japan, while Kuwana (2019) showed that patriotism is conducive to voting for the LDP in recent elections.
Based on a review of these studies, it appears pertinent to reconsider the relevance of authoritarian values in models of voting behavior in Japan. Indeed, the revival of nationalist and authoritarian rhetoric among right-wing elites suggests that the Japanese public is far from unresponsive to far-right appeals and that the effect of competing value preferences among voters may eventually unfold in the polls, thus conditioning voting behavior. Using survey data, our study examines the role of authoritarian values, which is a key ideological attribute of the far right (Mudde, 2007), in shaping the electoral preferences of Japanese respondents. Accordingly, we hypothesize that the variation in authoritarian values among the Japanese public has affected the probability of voting for the main parties since the beginning of the Koizumi era.

Data and methods
To assess the relevance of authoritarian values in explaining party preferences, we rely on data from the Asian Barometer Survey (ABS). 4 The ABS questionnaires in Japan have been administered in a face-to-face format consistently throughout the series to random electorates; therefore, we believe they constitute the most ideal data for our inquiry. 5 We combine five samples from the five ABS waves administered in Japan from 2003 to 2019, which we analyze using pooled multinomial logistic regressions with fixed effects where the vote for the LDP serves as the reference category. By pooling the five waves, we take advantage of the cross-temporal design of the ABS series, which allows us to run a comprehensive model encompassing the last two decades. In addition, we compare the explanatory power of authoritarian values and the other covariates in each wave separately. Our dependent variable is respondents' vote choice in five national elections including abstention. For the analysis, we selected the parties that contested all national elections since 2003 (Wave 1) and included the remaining parties in the category 'Others.' The four parties whose name and format did not change during the 2003-2019 period are the LDP, Komeitō or Clean Government Party (CGP), Social Democratic Party (SDP), and JCP. For the DPJ, which incurred several defections and eventually dissolved in 2016, we constructed an umbrella category 'DPJ/center-left' comprising the DPJ itself (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) and its successor parties in 2019 (Wave 5), namely, the Constitutional Democratic Party (CDP) and the Democratic Party for the People (DPP). 6 Our main independent variable is a measure of authoritarian values, which we derive from the ABS original battery 'Authoritarian/democratic values.' For the sake of consistency in our measurement, we selected only the items that were administered across all five waves. Next, we conducted separate reliability analyses for each wave using Cronbach's alpha to assess the degree of inter-consistency of the scales. For all sub-samples, the alpha coefficient is maximized when the first item is removed ('People with little or no education should have as much say in politics as highly-educated people'), indicating no substantial contribution of this item to the scales. With the removal of Item 1, we obtained a satisfactory fit of the measurements, with the alpha coefficients exceeding 0.7 for all waves. Finally, to compute our scale, we combined the items into a single construct by taking the mean score of the responses. Table 1 shows the statements used to construct the authoritarian values scale and the individual alpha coefficients for each wave.
Beginning with Adorno's (1950) seminal theory of the authoritarian personality, a number of studies have proposed different conceptualizations and measurements of individual authoritarian dispositions, among which Altemeyer's (1981) tripartite definition of right-wing authoritarianism (RWA), explained as the covariation of authoritarian submission, aggression, and conventionalism, has been widely accepted as a common framework. A noteworthy attempt to gage authoritarian personality traits among the Japanese public was made by Takano et al. (2020), who developed a Japanese version of Altemeyer's RWA using survey items adjusted to the Japanese context and found positive and significant correlations of their RWA measure and its two subdimensions (authoritarianism and conventionalism) with the LDP/CGP vote.
While acknowledging that measurements based on Altemeyer's RWA are not available for the time span of our inquiry, we believe that the factors included in the ABS batteries, namely, individual preferences for delegative forms of democracy, deference toward the established authority, and illiberal conceptions of the political community, address the core conceptual properties of authoritarianism 4 To investigate the relevance of authoritarian values in voting behavior, this study relies on observational data that allow us to gage changes and stability in the importance of this value dimension over time. This by no means excludes the possibility of rigorously testing its causal effect in an experimental setting in future studies. as conceived in other studies. Indeed, looking at the items utilized by Flanagan in his A-L scale (see Table 2), we find some theoretical affinity between the A-L statements and our measurement.
In particular, the first three items in Table 2 reflect similar dispositions of deference to authority as those captured by Items 2 and 7 from the ABS battery (see Table 1), and Flanagan's fourth item 'Cut discussions short when divided to avoid bad feelings' is consistent with the idea that a diversity of opinions endangers the harmony of the community (Items 2 and 8 in Table 1).
In addition to authoritarian values, we assess the effect of two other non-economic dimensions commonly associated with far-right discourses: national pride and anti-immigration. The emphasis on the uniqueness of Japanese identity and values as reviewed above suggests that Koizumi and Abe's reliance on nationalist appeals may reflect similar dispositions among LDP voters. The ABS project includes a survey item measuring the extent to which respondents feel pride in being Japanese. Though unavailable in the 2003 questionnaire (Wave 1), we include this predictor in a separate pooled model for Waves 2-5. Furthermore, to account for the effect of individual attitudes toward immigration, we estimate an additional model where we include a survey question in which participants are asked to express their views on immigration issues. Available for Waves 4 and 5 only, we expect that higher values indicating respondents' disagreement with increasing the inflow of immigrants would negatively correlate with voting for left-wing parties.
Next, we selected a set of additional predictors to control for individual socio-demographic characteristics, namely, respondents' age, gender, education, family income, rural or urban residence, and employment condition as an employee or self-employed. Furthermore, we relied on a survey item asking respondents to rate Japan's overall economic situation to assess the effect of individual evaluations of the economy on vote choice. To account for relevant factors commonly associated with political disengagement and reluctance to go to the polls, we controlled for individual interest in politics, which Mifune (2005) found to be a significant predictor of turnout in the 1990s. Finally, we measured respondents' degree of trust in institutions, which we also expect to negatively correlate with abstention, by constructing a scale of institutional trust using the items in the ABS battery 'Trust in institutions' (see the Technical appendix). Strongly agree (4), somewhat agree (3), somewhat disagree (2), strongly disagree (1) 1. People with little or no education should have as much say in politics as highly-educated people 2. Government leaders are like the head of a family; we should all follow their decisions 3. The government should decide whether certain ideas should be allowed to be discussed in society 4. Harmony of the community will be disrupted if people organize lots of groups 5. When judges decide important cases, they should accept the view of the executive branch 6. If the government is constantly checked [i.e. monitored and supervised] by the legislature, it cannot possibly accomplish great things 7. If we have political leaders who are morally upright, we can let them decide everything 8. If people have too many different ways of thinking, society will be chaotic Cronbach's alpha Wave 1 (2003) Wave 2

Results
This section presents the results of the statistical analyses in three pooled models, as outlined.
The results are displayed as the odds ratios of vote choice for the main parties and abstention relative to the vote for the LDP. Table 3 shows the results for Model 1, which covers the whole 2003-2019 period. In Model 2 (Table 4), we include national pride as an additional predictor starting from the second ABS wave. Model 3 (Table 5) assesses the impact of immigration issues in 2016 and 2019 (Waves 4 and 5). The odds ratios for the measure of authoritarian values suggest significant effects for all outcome groups with the exception of abstainers and the CGP. Overall, the estimated effects of authoritarian values on vote choice seem to indicate that individual differences on this value dimension structure vote preferences along the dividing line between the government and opposition camps. Higher scores on authoritarian values significantly decrease the probability of the non-LDP vote, with considerable negative effects in the case of the more left-wing opposition parties, the SDP and the JCP: for every additional unit of authoritarian values, the odds of voting for the SDP and the JCP decrease by approximately 69 and 80%, respectively. 7 Regarding the largest center-left party, the DPJ (CDP and DPP in Wave 5), our model shows that the odds of voting for the DPJ/center-left decrease by approximately 45% for every one unit increase in authoritarian values, suggesting less pronounced, though still significant, value differences with LDP voters. Not surprisingly, a comparison between vote choice for the LDP and its ally the CGP yielded no significant difference in voters' value positioning. Interestingly, though our estimate failed to reach the threshold of statistical significance, the direction of the relationship between authoritarian values and abstention is positive.
Looking at the odds ratios for the other covariates, our model reveals socio-demographic profiles similar to those identified by major studies of Japanese voting behavior and substantially confirm previous findings that LDP supporters are predominantly male and generally older than those of other parties (see Miyake, 1985;Kabashima, 1998). The odds ratios for rural residence suggest that respondents living in non-urban areas are more likely to vote for the LDP, a result that clearly reflects the party's traditional advantage in Japan's rural prefectures. As for respondents' education and income, LDP voters exhibit significantly higher levels of educational attainment than abstainers and CGP voters and tend to be generally wealthier than the other outcome groups.
In terms of respondents' employment condition, our analysis provides evidence that the jimaehi-jimae opposition between self-employed and hired workers (see Miyake, 1985) has relatively limited impact on vote choice. Besides the insignificant effect of self-employment, hired workers appear to be more likely to vote for the CGP, consistent with the fact that this party mainly garners the support of urban employees.
Next, the results for individual evaluations of the state of the economy suggest that more positive views of Japan's economy decrease the chances of abstention or voting for the DPJ/center-left and the JCP and, though statistically insignificant, the estimated odds ratios are lower than 1 for all other outcomes in the model.
As expected, whereas political interest reduces the probability of abstention, the opposite relationship holds for the DPJ and its successor parties: respondents who are more indifferent toward political matters are more likely to desert the polls, and those with higher degrees of political interest have a higher probability of voting for the DPJ, in line with what we may expect for the voters of a mainstream center-left party. Finally, our scale of institutional trust negatively correlates with all outcome categories, except for the CGP, where the estimated odds ratios are not statistically significant.
The broad picture emerging from Model 1 is that the variation in authoritarian values reflects voting patterns similar to those identified by Flanagan for the 1955 system. Indeed, the effect of authoritarian values on vote choices appears particularly evident when looking at predicted 7 Although the direction of relationship between authoritarian values and the vote for the SPD and JCP is negative as expected, this finding should be interpreted with caution given the relatively small number of cases in these categories.
probabilities. Figure 1 shows the predicted probabilities for each outcome category estimated from the logit coefficients in the model.
In Figure 1, the predicted probability curves for the LDP and abstainers exhibit a clear upward direction, as opposed to the DPJ/center-left vote, whose probability sharply reduces as respondents' Pooled multinomial logistic regression with fixed effects for each wave. ***P < 0.001, **P < 0.01, *P < 0.05. Source: Authors' elaboration from Asian Barometer Survey. scores on authoritarian values increase. Consistently with the regression results, the estimated marginal effects for the SDP and the JCP indicate almost zero probability of voting for these parties in the bottom-right region of the figure, where the scores on authoritarian values are higher than 3. As for the CGP, the relatively flat line suggests no consistent relationship between authoritarian values and the CGP vote. Next, we evaluate the impact of respondents' degree of national pride and individual attitudes toward foreign immigrants. Model 2 assesses the effect of national pride on the vote, adjusted for the covariates of Model 1. Model 3 includes anti-immigration as an additional predictor. The results are displayed in Tables 4 and 5, respectively.
In line with our expectations, Models 2 and 3 show that as respondents' degree of national pride increases, the odds of voting for the main left-wing parties reduce significantly. This finding suggests Pooled multinomial logistic regression with fixed effects for each wave. ***P < 0.001, **P < 0.01, *P < 0.05. Source: Authors' elaboration from Asian Barometer Survey. Pooled multinomial logistic regression with fixed effects for each wave. ***P < 0.001, **P < 0.01, *P < 0.05. Source: Authors' elaboration from Asian Barometer Survey.
that the LDP's emphasis on national identity and patriotism reflects greater attachment to the nation among right-wing voters as compared to voters on the left. 8 Likewise in Model 3, the estimated odds ratios for respondents' preference for tighter immigration policies are lower than 1 for all outcome categories, although significant effects of anti-immigration sentiments are found for the JCP only. As for our main explanatory variable, the results for Models 3 and 4 substantially confirm the relationships found in Model 1: authoritarian values significantly reduce the probability of voting for the main left-wing parties relative to voting for the LDP. Overall, the multinomial logit models presented in this study provide substantial empirical evidence that value differences along the scale of authoritarian values relate to vote choice in the 2003-2019 period. In terms of expected vs predicted relationships, our findings are consistent with the idea that the LDP under Koizumi and Abe was effective at attracting voters with relatively higher authoritarian dispositions as compared to voters of all left-wing opposition parties. Interestingly, these results seem to deviate from previous findings that the hawkish turn of the LDP under Abe hardly reflected an equivalent drift to the right among Japanese voters (see Jou et al., 2017). 9 Yet, it is worth noting that our logit models do not address the question of whether there has been an authoritarian turn in the electorate as a whole, an aspect we discuss in further detail in the next paragraphs. In addition, these (apparently) divergent findings may reveal the existence of discrepancies between how respondents perceive themselves as conservatives or progressives (as in Jou et al.'s analysis) and their actual ideological scores derived from multiple survey questions.
To further validate our findings and determine the extent to which these values contribute to explaining vote choice, we conducted a series of likelihood ratio tests where the contribution of each independent variable to the fit of the models was assessed separately. The tests were performed on both the pooled samples and each individual ABS wave, as shown in Table 6.
In Table 6, authoritarian values are a significant predictor of vote choice in each model and subsample, and they have the highest χ 2 score after age in Model 1. Our tests also suggest some degree of explanatory power of the respondents' socio-economic status and their evaluations of the economy. In addition to authoritarian values, the three other non-economic variables significantly contributing to the fit of the models are: interest in politics, trust in institutions, and national pride. Interestingly, the insignificance of anti-immigration in Model 3 suggests little evidence that respondents' preferences for stricter immigration rules affect their voting behavior. Despite this unexpected finding, our likelihood ratio tests fully support the idea that values are important predictors of vote choice in addition to respondents' socio-economic status and economic evaluations, and suggest a reconsideration of the role of non-economic issues in explaining the patterns of voting behavior in contemporary Japan. Yet, notwithstanding the relevance of these values for the overall time frame of our inquiry, it is worth noting that there may be significant cross-temporal variation in the presence and intensity of authoritarian values. This aspect, which is not fully captured by our logit models, deserves a specific assessment. Figure 2 shows the average scores on authoritarian values across the five ABS waves divided by party category and Figure 3 illustrates the average predicted probabilities for voters scoring above each sample mean on authoritarian values.  Consistently with Jou et al.'s (2017) finding that there is little evidence of a rightward turn among Japanese voters, the trends displayed in Figure 2 indicate that there has not been a sizable increase in authoritarian values since 2003. Rather, the overall mean values for each sample exhibit a slight downward direction starting from 2007. Nonetheless, looking at the LDP/CGP voters and the voters of left-wing parties, there is a clear widening gap after 2011 that reaches a peak in 2019. These divergent trajectories are substantially in line with our model estimates, as shown in Figure 3. Among respondents scoring high on authoritarian values, whereas the estimated probabilities tend to converge over the 2003-2011 period, the gap widens significantly in 2016 and remains substantially large in 2019. These trends clearly suggest that while authoritarian values may not be on the rise among the Japanese public, the variation in this value dimension nevertheless reflects specific patterns of voting that are consistent over time.

Robustness checks
In addition to the likelihood ratio tests presented above, we conducted a wide range of empirical evaluations as additional robustness checks. First, we estimated the same models as in Tables 3-5 using instead multinomial probit, an estimation technique that Horiuchi (2001) found to be better suited than multinomial logit for modeling vote choice. Next, we included dummy variables of partisan identification for each party category to account for the effect of the psychological link between respondents' party identification and their vote choice. Prominent in the socio-psychological model of the Michigan school, party identification constitutes a stable psychological shortcut for voters expressing clear partisan preferences, which Lachat (2015) found to reduce the effect of party-voter issue distance on vote choice. Further robustness checks were conducted by estimating models with unweighted data and specifying the CDP as the DPJ's unique successor party in Wave 5. These alternative model specifications, applied either individually or in combination, did not yield substantially different results with regard to the effect of authoritarian values on vote choice. For further details, see the Technical appendix.

Conclusion
This article has proposed an empirical reassessment of the role of values in models of voting behavior in Japan. It has been argued that the intensification of nationalist and authoritarian appeals among Japan's conservative elites (see Nakano, 2016) likely reflected differences in the value positions of the general public and that these differences ultimately affected the electoral behavior of Japanese voters. Using the ABS original battery 'Authoritarian/democratic values,' we measured respondents' authoritarian dispositions and employed our scale of authoritarian values to estimate pooled multinomial logit models of vote choice from 2003 to 2019. The results confirm that the variation in authoritarian values affects the probability of voting for the main parties, with the main right-wing party, the LDP, catering to voters with relatively higher levels of authoritarianism than the voters of main left-wing parties.
In addition to authoritarian values, we examined the effect of two other dimensions that typically characterize the far-right narrative: national pride and anti-immigration. We found that although higher degrees of national pride reduce the probability of the non-LDP vote, immigration issues seem to play only a marginal role in shaping vote choices.
Overall, the results suggest a considerable salience of non-economic issues in contemporary Japan, whereby the current voter alignments mirror the opposition between a left-wing camp displaying greater attachment to the values of liberal democracy and pluralism embedded in Japan's pacifist constitution and a right-wing camp placing more emphasis on the values of authority, hierarchy, and social order.
Since the emergence of successful green parties and the subsequent rise of populist radical right parties in Western Europe, a number of studies have focused on the salience of a new dimension of conflict in addition to the conventional left-right or state-market political divisions. Prominent in European research, this dimension has been characterized as new vs old politics (Taggart, 1996;Müller-Rommel, 1989), libertarian vs authoritarian (Kitschelt, 1994), or GAL/TAN values -Green-Alternative-Libertarian vs Traditional-Authoritarian-Nationalist - (Hooghe et al., 2002).
In Japan, where cases of successful green parties or populist radical right challenger parties are missing, this dimension of contestation has received relatively little academic attention in recent years. However, it is noteworthy that the relevance of values along the authority-liberty dimension was emphasized early by Flanagan in his examination of the determinants of partisan preferences in post-war Japan. In light of these considerations, our findings suggest that these patterns of voting are not a new phenomenon in Japan and provide substantial empirical evidence in support of Flanagan's framework.
Yet, our inquiry also suffers from important limitations, for the most part due to data availability. In particular, the time span of the analysis does not allow for a comparison between the pre-and post-Koizumi eras, and therefore leaves unanswered the question as to whether the relationships found in this study constitute an enduring feature of Japanese politics or whether the salience of these values was intensified by Koizumi and Abe's nationalist agendas. Indeed, as the ABS project does not include specific survey questions about leaders, we are not able to disentangle the effects of authoritarian values on respondents' reported vote for political parties from their specific preferences for individual leaders.
Despite these limitations, this study has important implications for our understanding of contemporary Japanese politics and makes significant contributions to the broader debate about populism in Japan (see Fahey et al., 2021). The empirical findings generally support the argument that the LDP's resistance to the international liberal order enabled it to address certain TAN value priorities of the more conservative parts of the Japanese public (see Lind, 2018), thereby reducing the risk of successful competitors on the right. Yet, notwithstanding the heterogeneity of values and positions among nonvoters, the relatively pronounced degree of authoritarian orientations among abstainers constitutes a novel finding that may challenge the idea that demands for authoritarian populism are already fulfilled by the current political supply, thus indicating the potential for the emergence of new challenger parties. Finally, our findings suggest that as far as these value positions are concerned and as long as the available party options remain unchanged, we may expect that an increase in the electoral participation in subsequent elections would be more beneficial for the LDP than for the opposition parties, a consideration that has important implications for the future trajectories of Japanese democracy.