As the evidence of party cueing is primarily obtained from studies conducted in the USA, which has a long history of a stable two-party system, it remains unclear whether the evidence is generalizable to political contexts outside the USA, especially to countries with multiparty systems. In fact, several noteworthy non-US studies have indicated that the effect of party cues on policy attitudes is limited (Merolla et al., Reference Merolla, Stephenson and Zechmeister2008, Reference Merolla, Stephenson, Zechmeister, Blais, Laslier and Van der Straeten2016; Kobayashi and Yokoyama, Reference Kobayashi and Yokoyama2018). Specifically, Kobayashi and Yokoyama (Reference Kobayashi and Yokoyama2018; K&Y hereafter) targeted Japan's unstable multiparty system and demonstrated that the effects of single-party cues are null in Japan, suggesting the limited generalizability of US-based findings, which consider partisanship to be the bedrock factor in the formation of policy preferences.
However, it would be hasty to conclude that Japanese voters do not utilize any cues in forming policy preferences from the findings of K&Y because there are other types of cues apart from single-party cues (Mondak, Reference Mondak1993). Although K&Y also found that the ideological leanings of major press organizations do not serve as effective cues for Japanese voters, other types of cues have been largely unexplored. In particular, the effect of political leader cues as substitutes for party cues warrants empirical investigation in Japan because attitudes toward prime ministers have played increasingly important roles in Japanese politics since Koizumi Junichiro formed his first cabinet in 2001 (Jou and Endo, Reference Jou and Endo2015). Drawing on this so-called ‘presidentialization’ of Japanese politics, the present study first examines the effect of prime minister cues as potential substitutes for party cues in Japan.
In addition to examining the effect of prime minister cues, this study articulates the effect of multiple-party cues that may also potentially substitute for the effect of single-party cues. K&Y demonstrated the null effect of party cues by examining the interaction effects between single-party cues in an experimental vignette and the respondents' party identification. However, the effect of party cues in a multiparty system may not arise from the availability of single-party cues, but rather from the information utility produced by the combination of multiple-party cues that vary across the ideological dimension. Multiple-party cues are defined as cues that indicate that a policy is supported by multiple rather than single ideologically distinct parties. As an illustration, when it is indicated that ideologically uncongenial parties such as the conservative Liberal Democratic Party (LDP) and the progressive Japanese Communist Party (JCP) both support a policy, voters may perceive a higher level of agreement among political elites, and thus greater legitimacy of the policy, compared with a case where only ideologically similar parties support it, because the information utility is enhanced when its sources are mutually independent (Harkins and Petty, Reference Harkins and Petty1981, Reference Harkins and Petty1987). In this example, Japanese voters are expected to show a stronger preference for the policy because the fact that the ideologically uncongenial LDP and JCP independently assessed the legitimacy of the policy and both still support it has greater information utility than knowing whether one or the other party does so. That is, in the context of a multiparty system, party cues are expected to serve as a more effective heuristic when presented in the form of multiple- rather than single-party cues.
Our online survey experiment clearly demonstrates that prime minister cues have a greater impact on Japanese voters' policy preferences than party cues, especially among the supporters of opposition parties. Furthermore, multiple-party cues are found to influence the policy preferences, unlike single-party cues.
1. Muted effect of party cues in Japan
Although democratic theories presuppose that voters possess political knowledge to form their own policy preferences, ordinary voters' knowledge levels are on average quite low (Delli Carpini and Keeter, Reference Delli Carpini and Keeter1996), and their policy preferences are loosely constrained at best (Converse, Reference Converse and Apter1964). However, by utilizing peripheral cues, voters with bounded rationality can compensate for a lack of political knowledge and form policy preferences in an ‘as-if rational’ way (Popkin, Reference Popkin1991; Schaffner and Streb, Reference Schaffner and Streb2002).
While cues take a variety of forms, those most readily available to voters are party cues (Lau and Redlawsk, Reference Lau and Redlawsk2001). As an illustration, Cohen (Reference Cohen2003) tested the effect of party cues on policy preferences by manipulating a party's issue position in a vignette on welfare policies. The results showed that when party identifiers are presented with a vignette indicating that their party endorses a policy, they support the policy even when it is inconsistent with their ideology. That is, when party cues are available, Democrats support even a stringent welfare program, while Republicans support even a lenient one, indicating the power of party cues in policy preference formation.
It should be noted, however, that most of the empirical evidence of party cueing is from the USA, which potentially overestimates the effect of party cues. The USA has a stable two-party system and the ideological positions of the Democratic and Republican parties are remarkably polarized (Abramowitz, Reference Abramowitz2010). Because ideological distinctions between parties magnify the effect of party cueing (Brader et al., Reference Brader, Tucker and Duell2013), the polarizing nature of modern US politics is expected to make tests of party cueing lenient. Therefore, it remains unclear whether party cues serve as effective shortcuts in policy preference formation in non-US contexts. As an illustration, Merolla et al. (Reference Merolla, Stephenson, Zechmeister, Blais, Laslier and Van der Straeten2016) demonstrated that in Canada, regardless of the level of political sophistication of the voters, providing candidate party cues is less effective for promoting particular voting choices than providing their issue positions, which demonstrates the limited power of party cues. By contrast, in the relatively new democracy of Brazil, Samuels and Zucco (Reference Samuels and Zucco2014) experimentally demonstrated that voters can form policy preferences that are consistent with their own partisanship only when party cues are available, suggesting that the impact of party cueing on policy preferences found in the USA is generalizable to Brazil.
Motivated by the potentially country-specific nature of party cueing, K&Y investigated the effectiveness of single-party cues in Japan by following the experimental design of Samuels and Zucco (Reference Samuels and Zucco2014). Specifically, K&Y provided the participants with a vignette that describes either the authorization of the use of collective self-defense (a high-salience issue) or the amendment of the temporary staffing services law (a low-salience issue) and manipulated the availability of single-party cues in the vignettes. The results indicated that the effect of single-party cues in forming policy preferences was limited, if it exists at all, in Japan. That is, regardless of the level of issue salience, the availability of LDP and Democratic Party of Japan (DPJ) cues did not significantly impact policy preferences.
In summary, because the role played by single-party cues in forming policy preferences is not necessarily paramount in countries outside the USA, more caution is required to generalize US findings on party cueing to Japanese political contexts. That said, it would be premature to conclude that Japanese voters do not utilize any cues in forming their policy preferences. Although K&Y demonstrated that the policy preferences of Japanese voters are not swayed by the availability of single-party cues, there are many other types of cues besides party labels that are available when voters form policy preferences. To gain more regionally relevant knowledge, the present study tests whether Japanese prime minister cues act as substitutes for party cues, and subsequently examines the effect of information utility produced by multiple- rather than single-party cues.
2. Presidentialization and political leader cues
The presidentialization of politics has been observed in Western democracies, including the USA and UK (e.g., Heffernan and Webb, Reference Heffernan, Webb, Poguntke and Webb2005; Webb and Poguntke, Reference Webb, Poguntke, Poguntke and Webb2005; Garzia, Reference Garzia2011), which arguably enhances the effect of political leader cues. Although there has been no consensus regarding the definition of presidentialization (Mughan, Reference Mughan2000), it commonly refers to the growing resources and autonomy of political leaders (Poguntke and Webb, Reference Poguntke, Webb, Poguntke and Webb2005; Garzia, Reference Garzia2011) and the increasingly important roles played by political leaders in electoral processes, including the larger weight that voters place on their impressions of leaders (e.g., McAllister, Reference McAllister, Dalton and Klingemann2007). In particular, Graetz and McAllister (Reference Graetz and McAllister1987) highlighted how the shift from party politics to media politics is driving presidentialization, whereby the media strongly shape the image of political leaders.
Political leaders are motivated to expose themselves in the media because of televised politics and the concurrent changes in the electoral strategies of parties (Garzia, Reference Garzia2011). As an illustration, the mass media increasingly make direct reference to political leaders rather than the parties to which they belong (Dalton et al., Reference Dalton, McAllister, Wattenberg, Dalton and Wattenberg2000). At the same time, political leaders strategically utilize the media to convey their own visuals and personalities and to attract public attention (Kavanagh, Reference Kavanagh1995). Increased exposure to leaders' visuals and personalities provides citizens with opportunities to form more personalized impressions of leaders (Grabe and Bucy, Reference Grabe and Bucy2009, Reference Grabe, Bucy, Bucy and Holbert2013). Consequently, citizens tend to evaluate political leaders by making use of the cognitive frameworks that are typically used to form personal impressions of others in daily life (Rahn et al., Reference Rahn, Aldrich, Borgida, Sullivan, Ferejohn and Kuklinski1990; Garzia, Reference Garzia2011), which particularly lowers the bar for less knowledgeable voters to make sense of politics based on personal impressions of leaders (Lentz and Lawson, Reference Lentz and Lawson2011).
Perceived images of political leaders through media exposure are expected to serve as a heuristic for voters to reduce their cognitive load when processing complicated policy information (Rahn et al., Reference Rahn, Aldrich, Borgida, Sullivan, Ferejohn and Kuklinski1990; Capara and Zimbardo, Reference Capara and Zimbardo2004). Therefore, along with the advancement of presidentialization, voters are more likely to form personalized impressions of political leaders, thus making it easier for them to extrapolate their likes and dislikes of leaders to candidates and parties (Funk, Reference Funk1999). In short, heuristic formation of policy preferences based on the personalized likes and dislikes of leaders will be more prevalent in presidentialized politics. Indeed, the effect of perceived personalities of political leaders on voting behavior is demonstrated to be stronger among those who consume TV news more heavily (Edelman, Reference Edelman1988; Schmitt and Ohr, Reference Schmitt and Ohr2000; Poguntke and Webb, Reference Poguntke, Webb, Poguntke and Webb2005).
3. Prime minister cues in Japan
Can political leader cues in Japan be an effective cognitive shortcut? Research has shown that the influence of leaders is limited compared with that of parties in Japanese politics (Massey, Reference Massey1975; Hayao, Reference Hayao1993). However, for the following reasons, presidentialization is also likely to be under way in Japan.
First, the weakening of party identification in Japan is expected to accelerate presidentialization. Japanese voters' party identification has weakened significantly owing to the party realignment triggered by political reform in the early 1990s. The proportion of voters identifying themselves as independent was 35% in the early 1990s, but it increased to 50% in 1995 after long-time competitors LDP and the Social Democratic Party of Japan (SDPJ) formed a coalition government. This trend continued, and since 2012, the overall proportion of independents has been about 60% (McElwain and Umeda, Reference McElwain and Umeda2012). Independent voters detached from parties tend to vote (or not vote) in a contingent way and easily swing from one party to another between elections (Matsumoto, Reference Matsumoto2006). Therefore, as the solid support base of political parties has been undermined – even that of the LDP, the catchall party in Japan – the relative importance of political leaders has risen, especially among nonpartisans.
In addition, the news coverage that increasingly focuses on the prime minister is assumed to facilitate presidentialization. Specifically, since the first Koizumi cabinet was formed in 2001, the Japanese news media have paid particular attention to political leaders (Osaka, Reference Osaka2014). As an illustration, Maeda and Hirano (Reference Maeda and Hirano2016) examined trends in the total amount of time devoted to media coverage of the prime minister on TV from 2007 to 2012. They found that annual TV airtime for prime ministers increased by 1,449 hours. Therefore, today's Japanese voters are exposed to the prime minister on TV to an unprecedented level.
As a consequence of presidentialization, attitudes toward political leaders are playing increasingly important roles in Japanese politics. By examining longitudinal survey data from the 1980s to the 2000s, Jou and Endo (Reference Jou and Endo2015) demonstrated that the impact of the feeling thermometer scores of political leaders on voting behavior is growing. In particular, as prime minister cues are the most readily available leader cues, they are expected to serve as an effective shortcut when Japanese voters form their policy preferences. Furthermore, because the effect of leader cues is especially pronounced when the leaders are from conservative or catchall parties (Lobo, Reference Lobo2008), the effect of prime minister cues would be especially prominent when the prime minister is from the LDP.
Based on the above arguments, the present study first examines the effectiveness of prime minister cues. As the effect of leader cues depends on whether a respondent supports their party (Nicholson, Reference Nicholson2012), we examine the heterogeneity of prime minister cueing between the supporters of incumbent and opposition parties. As Japanese politics at the time of this study were structured as a competition between the incumbent coalition of the LDP and Komeito and the opposition DPJ and JCP in electoral cooperation, prime minister cues work as in-party leader cues for the supporters of incumbent parties and as out-party leader cues for supporters of opposition parties. Although Nicholson (Reference Nicholson2012) found no effect of in-party leader cues on policy preferences, the present study examines the effect of prime minister cues on both incumbent and opposition supporters because other research indicates that in-group leaders are highly persuasive (Hogg, Reference Hogg2001) and in-group bias is more potent than out-group bias (Brewer, Reference Brewer2007).
H1:
Supporters of the incumbent parties approve more strongly of a policy when it is indicated that the prime minister supports it.
H2:
Supporters of the opposition parties disapprove more strongly of a policy when it is indicated that the prime minister supports it.
In addition to examining the effectiveness of prime minister cues in forming policy preferences, we pit the effect of prime minister cues against single-party cues. Nicholson (Reference Nicholson2012) found that the effect of leader cues on policy preferences is larger than that of party cues because leader cues are more personalized, so they elicit stronger effective responses than party cues. As the prime minister in this study (PM Abe) is from the LDP, we compare the effects of prime minister cues with LDP cues. Furthermore, if the effect of Japanese party cues is found to be weak or null, this would replicate the results of K&Y, giving more credence to the finding. Therefore, the following hypotheses are proposed.
H3:
Supporters of the incumbent parties approve more strongly of a policy with indications of prime ministerial support than with indications of LDP support.
H4:
Supporters of the opposition parties disapprove more strongly of a policy with indications of prime ministerial support than with indications of LDP support.
4. Effect of multiple-party cues
Although K&Y found single-party cues to be ineffective in Japan, information about parties that support a policy is theoretically expected to convey valuable information to voters even in the midst of presidentialization because Japan's electoral system has a parliamentary system as well as proportional representation. What potentially reconciles this apparent inconsistency is the effect of multiple-party cues. Indeed, experimental studies in social psychology have indicated that multiple cues have greater information utility, thereby affecting attitudes more effectively than single cues.
Previous research on ‘multiple source effects’ in social psychology has indicated that even when the same argument is presented, the message is scrutinized more deeply and thus becomes more persuasive when it is delivered by multiple sources than by a single source. Using the issue of introducing ‘Senior Comprehensive Exams’ to the university curriculum, Harkins and Petty (Reference Harkins and Petty1981) showed participants a video in which one or three students presented their opinions on an issue to convince their audience, after which, the issue attitudes of the participants were measured as the dependent variable. There were three conditions in terms of the presentation of arguments in favor of the exams in the video: (a) one student presented three arguments (one source/three arguments); (b) each of the three students presented one argument (three sources/one argument); and (c) each of the three students presented three arguments (three sources/three arguments). The results indicated that the persuasion effect was stronger in the three-source/one-argument condition than in the one-source/three-argument condition, suggesting that multiple sources are more persuasive than single sources even when they convey the same number of arguments. Following this finding, Harkins and Petty (Reference Harkins and Petty1987) further showed that multiple sources are more persuasive because the perceived informational independence between multiple sources increases the perceived utility of the information.
On the basis of these findings from social psychology, it is predicted that the effect of party cues would be magnified when there are indications that multiple parties, rather than a single party, support a policy because the perceived informational utility of multiple-party cues would be larger than that of single-party cues. In particular, the effect of multiple-party cues would be larger when ideologically distinct parties support the same policy because the perceived information utility becomes higher when the sources are perceived to have reached their issue positions independently. Hence, the following hypothesis is proposed.
H5:
People more strongly favor a policy when there are indications that ideologically distinct multiple parties support it compared with when no party cues are available.
5. Method
5.1 Experimental procedure
Japanese adults aged 20–69 years were recruited from the online panel of Nikkei Research, a leading online survey firm in Japan. We solicited participation for the online experiment from 22,634 people registered in the panel, 2,047 of whom met the demographic criteria (i.e., Japanese nationality and age range) and agreed to participate in the study. From the 2,047 potential participants, we randomly sampled 1,800 for the study.
After measuring covariates, the experimental vignettes were presented. The 1,800 participants were assigned to one of the eight conditions presented in Table 1 in a completely random manner. We selected ‘the basic law for urban agriculture promotion’ as an issue to be used in the vignette because all parties approved this bill at the 189th ordinary session of the Diet and the salience of the issue was low, making cue-based shortcuts more likely (Arceneaux, Reference Arceneaux2008). After presenting the vignettes, the participants' policy preferences on the issue were measured as the dependent variable.
Table 1. Experimental conditions
Of the 1,800 participants, 1,134 completed the experiment. To examine attrition bias, we estimated a logit regression model predicting attrition on the covariates, including age, sex, education, the treatment, and interaction terms between the treatment and covariates as independent variables (see Table A1 in Online Appendix). The treatment, covariates, and their interaction terms were not significant, suggesting that the attrition was not associated with the treatment.
5.2 Measurement of covariates
In addition to the demographic variables, political interest, cabinet approval, feeling thermometer scores for each party, self-identified ideology, perceived ideological position of parties, and media exposure were measured as covariates (see Appendix 1 in Online Appendix for details of the measurement of covariates).
K&Y measured party identification as long-term partisanship (Taniguchi, Reference Taniguchi2012) to see how it moderates the effect of party cues. Although this measurement is specifically localized to the Japanese political context, the proportion of party identifiers in the sample is extremely small, except for the LDP when the party support is directly measured, because most of the electorate is independent. Dropping these independents from the analysis would significantly reduce the sample size, making the study underpowered to detect the effect of party cueing. Therefore, to classify the participants into supporters of incumbent and opposition parties while keeping as many of those who do not identify with parties in the sample as possible, the present study uses feeling thermometer scores of parties instead of self-reported party identification to operationalize party support. Specifically, the participants were classified as incumbent supporters when the feeling thermometer score of the LDP or Komeito was higher than those for both the DPJ and JCP. Conversely, the participants were classified as opposition supporters when the feeling thermometer scores for the DPJ or JCP exceeded those for both the LDP and Komeito. Operationalizing party support with feeling thermometer scores allows us to address the limited number of party identifiers because of weak partisanship and to boost the efficiency of the statistical analyses. Those with missing values in the feeling thermometer scores were dropped from the analyses.Footnote 1 Even when there were no missing values in the feeling thermometer scores, the participants were dropped from the analyses if their scores for incumbent and opposition parties did not exceed the scores of the two parties in the other group. For instance, those who had the same level of feeling thermometer scores for all four parties were dropped because they could not be classified into either incumbent or opposition supporters. As a result, data for 780 participants were used in the following analyses, of which, 60% were supporters of incumbent parties.
5.3 Vignettes
A brief vignette was presented in which the presence of party or prime minister cues was manipulated. The vignette of the control condition in which no party or prime minister cues were provided read as follows.
Control condition
As a policy aimed at promoting agriculture around the urban area, the basic law for urban agriculture promotion was submitted to this ordinary session of the Diet. Do you approve or disapprove of this bill?
Party cue conditions
The participants assigned to the single-party cue conditions (conditions 2–4 in Table 1) were presented with an additional description indicating that either the LDP, DPJ, or JCP supported the bill. Those assigned to the multiple-party cue conditions (conditions 5–7 in Table 1) were presented with an additional description indicating that one of the following three combinations of parties supported the bill: ‘LDP & DPJ’; ‘LDP & JCP’; or ‘LDP, DPJ, & JCP.’
As a policy aimed at promoting agriculture around the urban area, the basic law for urban agriculture promotion was submitted to this ordinary session of the Diet and (LDP, DPJ, or JCP), (LDP & DPJ, LDP & JCP, or LDP, DPJ, & JCP) support(s) the bill. Do you approve or disapprove of this bill?
In the prime minister cue condition, the party name was replaced with that of Prime Minister Abe.
Prime Minister Abe cue condition
As a policy aimed at promoting agriculture around the urban area, the basic law for urban agriculture promotion was submitted to this ordinary session of the Diet and Prime Minister Abe supports the bill. Do you approve or disapprove of this bill?
As a result, we set up eight experimental conditions, including a control condition, six conditions with party cues, and one condition with prime minister cues (Table 1).
5.4 Measurement of the dependent variable
The policy preference on the basic law for urban agriculture promotion was measured on a four-point scale ranging from ‘agree’ to ‘disagree.’ This measurement was linearly transformed into an approval rate ranging from 0 to 100 (M = 61.32, SD = 23.0).
Although our experimental design followed that of Samuels and Zucco (Reference Samuels and Zucco2014) and K&Y to maximize its comparability, it differed from these studies in the following respects. First, we did not present brief descriptions of the pros and cons of the policy to avoid in-depth processing of the information through the central route, which would potentially reduce reliance on cues. Second, unlike K&Y, we did not use the instructional manipulation check (Oppenheimer et al., Reference Oppenheimer, Meyvis and Davidenko2009) to exclude survey satisficers. This was intended to avoid obtaining a sample that overrepresented those who prefer deep processing of information rather than cue-taking when they form policy preferences (see K&Y for an explanation of why excluding satisficers can work against detecting the effect of cues).
6. Results
6.1 Covariate balance
Before estimating treatment effects, we tested covariate balance among the valid respondents (N = 780). Although a statistically significant imbalance was found in the frequency of reading Asahi Shimbun (F (7, 772) = 2.10, p < 0.05), the overall covariate balance was maintained (see Tables A2 and A3 in the Online Appendix).
6.2 Analysis
To test the effect of prime minister cues and multiple-party cues most efficiently, an ordinary least-squares regression predicting posttreatment policy preference on the basic law for urban agriculture was estimated with the treatment (control condition as the baseline), party support (0: incumbent parties; 1: opposition parties), and their interaction term as independent variables (Models 1 and 2 in Table 2). Model 1 estimates the treatment effect without the covariates. Model 2 additionally includes the frequency of reading Asahi Shimbun (the imbalanced covariate), sex, age, and cabinet approval as covariates to boost the efficiency of estimating treatment effects. To test Hypothesis 5, we also estimated the average treatment effect without including interaction terms (Models 3 and 4 in Table 2). Model 3 estimates the average treatment effects without the covariates; Model 4 estimates them with covariates. Based on the estimated model with covariates (Models 2 and 4 in Table 2), the point estimates and their 95% confidence intervals were plotted for each combination of the treatment (Figures 1 and 2).
Figure 1. Treatment effects of LDP and prime minister cues on policy approval rate.
Figure 2. Treatment effects of multiple-party cues on policy approval rate.
Table 2. Treatment effect of party and leader cues on policy approval rate
Standard errors in parentheses.
***p < 0.001, **p < 0.01, *p < 0.05, †p < 0.1.
6.3 Effect of prime minister cues
Figure 1 illustrates the effect of prime minister cues based on Model 2 in Table 2. No treatment effect was found among the supporters of incumbent parties, providing no evidence to support H1. By contrast, the treatment effect was statistically significant among the supporters of opposition parties. The approval rate of the opposition supporters assigned to the prime minister cue condition was 49.07%, while that of those assigned to the control condition was 58.58%. To compare the mean differences between the two conditions, we performed a linear combination test. Specifically, the delta measure was obtained by subtracting two regression coefficients for the prime minister cue condition and the control condition and subjected to a t-test. The difference in the approval rate of the policy between the two conditions was 9.51%, which was statistically significant (t = 2.03, p < 0.05). This result supports H2.
Next, to test H3 and H4, the effect of prime minister cues was compared with that of LDP cues based on Model 2 in Table 2. The result indicated that the effect of prime minister cues was not statistically distinguishable from that of LDP cues among the supporters of incumbent parties (LDP condition = 65.95 vs. prime minister condition = 63.51, t = 0.58, n.s.). Therefore, H3 is rejected. By contrast, among the supporters of opposition parties, those assigned to the prime minister condition demonstrated a significantly lower level of approval than those assigned to the LDP condition (LDP condition = 59.57 vs. prime minister condition = 49.07, t = 2.08, p < 0.05). That is, although the prime minister is the president of the LDP, prime minister cues alienate the supporters of opposition parties more strongly than the cue of his party, which supports H4.
In summary, prime minister cues turned out to be ineffective in forming policy preferences of the incumbent supporters, suggesting that prime minister cues do not work as the in-group leader cue. However, prime minister cues strongly shaped the policy preference of opposition supporters. These results are consistent with the findings that political leader cues are particularly effective in forming negative policy preferences when they are presented as out-group cues (Nicholson, Reference Nicholson2012). Although limited among the opposition supporters, the results indicated that the prime minister cue was more influential than the LDP cue, suggesting that leader cues matter more than corresponding single-party cues in Japan.
6.4 Effect of multiple-party cues
To test H5, two multiple-party cue conditions, LDP & JCP cues and LDP, DPJ, & JCP cues, were compared with the control condition. We focus on these two multiple-party cue conditions because we expect information utility to increase as the number of the parties increases in the cue. LDP & DPJ cues were not the focus of the analysis because the ideological distinctiveness between LDP and DPJ is expected to be smaller than that between the LDP and JCP. In the following analyses, however, the results of all multiple-party cues are presented for transparency.
To test H5, the average treatment effects of multiple-party cues were tested. As Models 3 and 4 in Table 2 indicate, the LDP & JCP condition and the LDP, DPJ, & JCP condition showed significantly higher approval of the policy compared with the control group. However, as expected, the LDP & DPJ condition did not show such an effect, which suggests that multiple-party cues are effective not because they provide information on the merits of a policy by showing that many politicians endorse it, but because they indicate that ideologically incompatible parties nevertheless agree to support the same policy. In summary, the multiple-party cues consisting of ideologically uncongenial parties boosted overall approval of the policy, lending support to H5. The approval rates of the LDP & JCP cue condition and the LDP, DPJ, & JCP cue condition were not statistically distinguishable.
7. Discussion
Motivated by K&Y, who showed a null effect of single-party cues in Japan and the increasing presidentialization of Japanese politics, the present study tested whether prime minister cues in place of party cues help Japanese voters form policy preferences. Furthermore, to test the effect of party cues that are unique to a multiparty system, we investigated the effect of multiple-party cues, indicating the policy is supported by multiple ideologically distinct parties, rather than a single party.
The results showed that while prime minister cues are not utilized by the supporters of incumbent parties, those of opposition parties demonstrated significantly less approval of the policy when it was indicated that the prime minister supported the policy. That is, although prime minister cues do not help form policy preferences as in-group leader cues, they certainly work as out-group leader cues for opposition supporters to form negative policy preferences.
The result that prime minister cues are effective only among opposition party supporters is likely attributable to the political confrontation between the incumbent and opposition parties in Japan. Nicholson (Reference Nicholson2012) argued that when political parties confront each other, their supporters are motivated to distinguish themselves from the opposing party by forming negative attitudes toward policies that are approved of by the out-group leaders. Therefore, given the confrontational nature of current Japanese politics described above, the effectiveness of prime minister cues can be attributed to PM Abe's status as an out-group leader for the supporters of opposition parties.Footnote 2 Furthermore, while the effects of LDP and prime minister cues were both null and indistinguishable among the incumbent party supporters, prime minister cues showed stronger negative impacts than LDP cues on policy preferences among the opposition party supporters. That is, the opposition party supporters disapproved of the policy not because it was supported by the LDP, but because it was supported by PM Abe, suggesting that leader cues are potent in Japan.
Another novel finding of the present study, which adds to the literature on party cueing, is the effect of multiple-party cues. That is, a cue indicating that ideologically distinct parties support a policy enhances the approval of the policy (H5). These results are consistent with a series of findings that endorsements from multiple information sources enhance the persuasiveness of a message (e.g., Harkins and Petty, Reference Harkins and Petty1981, Reference Harkins and Petty1987; Moore et al., Reference Moore, Mowen and Reardon1994). It is possible that Japanese voters take multiple-party cues as a sign of the legitimacy of a policy – ‘If ideologically incompatible parties agree to support the policy, it must be because the policy has a lot of merit.’
The present study provides several insightful implications. First, given the effectiveness of prime minister cues as out-group leader cues, and PM Abe's increasing media presence because of the presidentialization of Japanese politics (Krauss and Nyblade, Reference Krauss and Nyblade2005), the incumbent parties and politicians will face greater difficulty in gaining support from opposition supporters for the policies they propose. This tendency could lead to the polarization of policy preferences between incumbent and opposition supporters because, as the leadership of the prime minister is strengthened, opposition supporters will oppose the policies he endorses just because they dislike him. As an illustration, PM Abe, who is seeking a constitutional amendment, repeatedly expresses his hope that the Japanese public will consider this issue. However, in light of the findings of this study, the more active the leadership that PM Abe shows on the issue of constitutional amendment, the more likely opposition supporters are to use their negative sentiments toward the prime minister as a shortcut to oppose the amendment, making careful deliberation even less likely. Furthermore, to the extent that Japanese politics is presidentialized, it is possible not only that those without party identification might be able to form policy preferences based on the personalized evaluation of the prime minister, but also that even the public's support for the party represented by the prime minister may be dependent on the personalized likes and dislikes of him or her. To investigate the effects of prime minister cues more rigorously, the feeling thermometer score of the prime minister needs to be measured in future studies. A significant interaction term between the prime minister cue and his/her feeling thermometer score would demonstrate that the prime minister cue is effective, even for those who do not have clear party identification.
Second, the finding that multiple-party cues increased support for the policy provides a clue as to how a broader consensus on policies may be built. In fact, in the Japanese Diet, quite a number of less confrontational bills are approved by all parties unanimously through customary cooperation between the incumbent and opposition parties (Masuyama, Reference Masuyama2001; Kawato and Masuyama, Reference Kawato and Masuyama2005). In this situation, the multiple-party cues indicating that ideologically distinct parties approved of the policy are expected to work as a heuristic to positive policy preferences, broadening support from the public. Even when the incumbent and opposition parties sharply disagree on a policy, polarization of opinion may be avoided by stressing that both camps agree on the general goals of the policy, despite their disagreements on the details, making the multiple-party cues more effective in gaining broader support for the policy.
Several limitations remain to be addressed. Most importantly, although the present study operationalized the incumbent and opposition supporters by using the feeling thermometer scores of parties to maximize the sample size used in the analyses, there was still a substantial number of ‘don't know’ responses to the feeling thermometer scales. In addition, there were substantial numbers of independents who could not be identified either as incumbent or as opposition supporters (e.g., they had the same feeling thermometer scores for all parties). As a result, 31.21% of the participants were dropped from the analysis, which reduced its statistical power. However, it should be noted that the statistical power was significantly improved compared with the K&Y study, in which conventional measurement of party identification was used, leading to 61.46% of the participants being dropped from the analyses. In fact, post hoc power analyses demonstrated that the observed effect size (Cohen's f) was 0.23 for the interaction terms between treatments and party support (Model 1 in Table 2) and 0.16 for the average treatment effects (Model 3 in Table 2). These observed effect sizes are larger than the minimum detectable effect size of 0.14 (α = 0.05, power = 0.8, n = 780). Therefore, our experimental design has sufficient statistical power to detect the treatment effects.
To address the issue of ‘don't know’ responses, we also employed multiple imputation for missing data on feeling thermometer scores to check the robustness of our findings and found that the results were similar to those without imputation (see Table A4 in Online Appendix). Indeed, the issue of ‘don't know’ responses directly speaks to the implication of this study. Under the current unstable Japanese party system where ‘don't know’ responses are prevalent in the feeling thermometer scores of parties, the effects of single-party cues on policy preferences are likely to be muted; in contrast, the utility of the prime minister cue will be relatively enhanced. At the same time, as newer parties' feeling thermometer scales would elicit more ‘don't know’ responses, the effect of multiple-party cues consisting of a combination of parties with longer histories such as the LDP and JCP will be increasingly large. To validate further our operationalization using feeling thermometer scores, future studies should replicate our findings by using the conventional measurement of party identification.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1468109919000021
Tomoya Yokoyama is an assistant professor of Department of Communication and Media Studies, Rikkyo University. His research focuses on political communication and political behavior.
Tetsuro Kobayashi is an associate professor of Department of Media and Communication, City University of Hong Kong. His research focuses on political communication and psychology in East Asian countries.
