2. Theory and hypothesis

Multiple studies claim that individual-level political ideology is associated with opposition to intergroup apologies. For example, Hornsey et al. (Reference Hornsey, Schumann, Bain, Blumen, Chen, Gómez and Wohl2017) demonstrated that conservative ideologies negatively correlated with one's general tendency to apologize and this relationship was mediated by entity belief and social dominance orientation (SDO), where SDO emerged as a more robust mediator variable. SDO is defined as one's tendency to endorse the existence of hierarchies among social groups (Pratto et al., Reference Pratto, Sidanius, Stallworth and Malle1994). Those who are high in SDO tend to be unwilling to apologize, even when they have committed transgressions. Extending this finding to the political apology context, Mifune et al. (Reference Mifune, Inamasu, Kohama, Ohtsubo and Tago2019) demonstrated that both conservative ideology and SDO were positively correlated with opposition to the Japanese government tendering apologies to Asian countries. Mifune et al. (Reference Mifune, Inamasu, Kohama, Ohtsubo and Tago2019) also found a positive correlation between opposition to governmental apologies and militarism, which is operationally defined as one's endorsement of ‘the use of military force sometimes to protect national interests in international politics.’

However, in three of the four studies by Karunaratne and Laham (Reference Karunaratne and Laham2019), political ideology was not found to correlate significantly with opposition to the US government's apologies after controlling for SDO and for right-wing authoritarianism (Altemeyer, Reference Altemeyer1998), while SDO continued to be significantly (though weakly) correlated with opposition to political apologies. Further, in the studies of Hornsey et al. (Reference Hornsey, Schumann, Bain, Blumen, Chen, Gómez and Wohl2017) and Mifune et al. (Reference Mifune, Inamasu, Kohama, Ohtsubo and Tago2019), SDO was relatively more strongly associated with resistance to apology than was ideology. In summary, the association between SDO and opposition to political apologies is robust, but the relationship with political ideology is not always consistent across studies. As summarized by Jost (Reference Jost2021), the study of ideology in psychology assumes that people share naive ideological conflicts, such as whether to accept or reject the status quo and whether to tolerate or reject disparities, and it is on this assumption that group apologetics research in psychology is conducted. However, as political science has long made clear, even if people do have ideologies, it is also true that it is difficult for many ordinary voters to link their ideological positions with attitudes toward policy issues. The weak relationship between political ideologies and opposition to intergroup apologies might be related to the fact that only a limited portion of people can view various political issues in light of their political ideology. Ideological divides such as left–right and conservative–liberal are concepts that experts frequently use to describe politics. However, political science literature has already established that, in general, voters do not have a solid and coherent understanding of ideologies. Converse (Reference Converse and Apter1964) found that political elites (e.g., politicians and political scientists) connect and understand a wide variety of political issues based on ideology, while most ordinary voters do not use this perspective. He conducted a study on the belief systems of voters using open-ended responses from American National Election Studies, finding that only 2.5% of the population understood politics based on ideology, and that this figure only reached 15% even when he relaxed the criteria for the acceptable level of understandings of political ideology. Furthermore, he demonstrated that, because there was no solid link between ideology and political issues in the belief system of a majority of citizens, there was little systematic association among the voters' attitudes toward separate political issues. The findings presented by Converse were harshly criticized, in part because they had the potential to undermine the very foundation of democracy, which is based on the premise that voters cast their ballots based on political issues (e.g., Bennett, Reference Bennett1973; Nie and Andersen, Reference Nie and Andersen1974; Pierce and Rose, Reference Pierce and Rose1974; Achen, Reference Achen1975; Miller and Miller, Reference Miller and Miller1976); however, the basic findings have yet to be disproved (e.g., Luskin, Reference Luskin1987; Kinder, Reference Kinder, Gilbert, Fiske and Lindzey1998; Lupton et al., Reference Lupton, Myers and Thornton2015). Moreover, the idea that many voters do not understand politics on the basis of ideology has been demonstrated not only in the USA, but also in Japan (Miyake et al., Reference Miyake, Kinoshita and Aiba1967; Miwa, Reference Miwa2015).

It has been suggested that even when voters appear to understand ideology, they may just respond to cues offered by the elites. Kinder and Kalmoe (Reference Kinder and Kalmoe2017), for example, pointed out that apparent ideology-based responses to social surveys may, in fact, reflect respondents' use of cues, such as those of political parties, in forming their political opinion. Moreover, with regard to the issue of whether to support a big government or a small one, Johnston et al. (Reference Johnston, Lavine and Federico2017) found that people with little political knowledge or concern appear to have rational attitudes based on self-interest, whereas people with political knowledge and interests appear to be irrational. However, this apparent irrationality is owed to their understanding of the issues in terms of conflicts of opinion between political elites. These results indicate that very few people perceive politics according to ideology. Accordingly, if only this limited fraction of the population has a coherent understanding of political ideologies, it is not surprising that correlations between ideology and political issues may depend on situational factors of salience and framing by the political elites.

Luskin (Reference Luskin1987, Reference Luskin1990) conceptualized political sophistication as the extent to which voters, like political elites, understand the various cognitive elements involved in politics by linking them together under an abstract framework of political ideology. According to the studies, the level of political sophistication is operationalized by the number and range of cognitive elements (e.g., knowledge of policies, political parties, or candidates) in people's belief systems, as well as the strength of the association between cognitive elements. Because political sophistication is defined by a combination of these multiple factors, there is no single measure that satisfies them all, and a variety of measures have been used. The most widely used are quiz-style political knowledge items that correspond to the number and range of cognitive elements in people's belief systems (e.g., Baum and Jamison, Reference Baum and Jamison2006; Weisberg and Nawara, Reference Weisberg and Nawara2010; Lau et al., Reference Lau, Patel, Fahmy and Kaufman2014; Bergbower et al., Reference Bergbower, McClurg and Holbrook2015; Rapeli, Reference Rapeli2018). As summarized by Delli Carpini and Keeter (Reference Delli Carpini and Keeter1996), political knowledge is an understanding of (a) the systems under which politics operates, (b) the politicians and political parties that currently play a role in politics, and (c) specific policy-related events. Of course, having knowledge of political systems, political actors, and current events is not the same as connecting various issues based on ideology. But understanding these issues is necessary for understanding the arguments of political elites, and is also closely tied to the interconnectedness of cognitive elements based on ideology (Luskin, Reference Luskin1987). This study does not claim that people do not have an ideology. However, we believe that conducting research based on the assumption that most people use the same axis as do political elites to define their own ideological positions and link themselves to specific policies hinders understanding of the relationship between ideology and opposition to collective apologies. Indeed, Jost (Reference Jost2021) also acknowledges that many people cannot understand policy issues based on the ideology shared by political elites, but argues that this does not necessarily mean that people do not have an ideology.

Thus, when investigating the association between resistance to a collective apology and ideology, we need to explore the possibility that the strength of this association differs by political knowledge. We hypothesize that political knowledge moderates the association between ideology and opposition to governmental apologies: in particular, the correlation between ideology and opposition is likely to be stronger for individuals with higher levels of political knowledge than for those with lower levels.

Intergroup apologies used in this study involve apologies made by the Japanese government for the atrocities committed by the Japanese military against Asian countries before and during World War II. Previous studies conducted in Japan have repeatedly demonstrated that conservatives show a more negative attitude toward intergroup apologies by the government, because evaluations of the old political system (i.e., the political system in the pre-World War II period) are closely connected to ideological confrontations between conservatives and progressivesFootnote ³ (e.g., Kabashima and Takenaka, Reference Kabashima and Takenaka1996, Reference Kabashima and Takenaka2012; Jou and Endo, Reference Jou and Endo2016a, Reference Jou and Endo2016b; Mifune et al., Reference Mifune, Inamasu, Kohama, Ohtsubo and Tago2019). That is to say, while Japanese progressives who disavow the prewar political system do not oppose apologies for shameful past misdeeds, conservative groups that do not necessarily disavow the prewar political system oppose apologies, possibly because they tend to imply the wrongness of the prewar political system. However, given the findings in ideological studies since the work of Converse (Reference Converse and Apter1964), discussed earlier, it is possible that these relationships merely reflect conflicts among those with political knowledge. In this study, we also address the relationship between SDO and resistance to collective apology in order to compare our findings that the relationship between ideology and resistance to collective apology differs depending on political knowledge. Hornsey et al. (Reference Hornsey, Schumann, Bain, Blumen, Chen, Gómez and Wohl2017) consider that SDO, which measures people's attitude toward between-groups inequality and equality, can be conceptualized as an ideological belief. However, we believe that there are significant differences between typical measure of political ideology and SDO. As noted above, a typical measure of political apology asks respondent to identify their own position on the unidimensional ideological axis and link it to political issues, which requires specialized knowledge of politics. The measure of SDO asks respondents to report their endorsement for intergroup status differences (Pratto et al., Reference Pratto, Sidanius, Stallworth and Malle1994), and evaluating political issues based on their level of endorsement does not require knowledge of politics. Moreover, a twin study revealed genetic influences on not only SDO but also concrete political attitudes toward issues related to intergroup status differences (Kleppestø et al., Reference Kleppestø, Czajkowski, Vassend, Røysamb, Eftedal, Sheehy-Skeffington, Kunst and Thomsen2019). The presence of genetic influences on both abstract (i.e., SDO) and concrete (political attitudes) levels is consistent with the idea that detailed knowledge of politics is not necessary to apply one's level of SDO to political issues. Thus, it is expected that the relationship between SDO and opposition to collective apology does not depend on political knowledge, whereas the relationship between political ideology and opposition to collective apology is likely to depend on the knowledge.

Furthermore, study 2 examines two factors that are associated with resistance to governmental apologies, both of which have little to do with political knowledge: general attitudes toward apologies and general attitudes toward neighboring countries that were affected by Japan's colonial policies in the past. Whether one approves of the act of apology itself and whether one favors the target country can be assumed to be conceptually independent of ideology. After adding independent variables to the regression analysis with the resistance to apology as the dependent variable, we will test whether the relationship found in study 1 between ideology, SDO, and resistance to group apology is replicated.

3. Study 1

3.2. Results

We first examined the reliability of the three items of resistance to governmental apologies, each corresponding to the three historical misdeeds (i.e., colonial and occupational rule, comfort women, and the massacre of Korean people in the aftermath of the 1923 Kanto Earthquake).Footnote ⁶ Cronbach's α coefficient was 0.89. Therefore, following Mifune et al. (Reference Mifune, Inamasu, Kohama, Ohtsubo and Tago2019), a single score of resistance to group apology was obtained by aggregating these three item scores. In particular, the three resistance scores were standardized within each item; thereafter, the three standardized scores were summed for each participant. Cronbach's α coefficient of the 16 items measuring SDO was 0.86, indicating sufficient internal consistency. Therefore, the average value of the 16 items was obtained and used as SDO.

For the 10 items measuring political knowledge, correct answers and incorrect answers (which included ‘I don't know’ responses) were assigned 1 and 0, respectively. Applying item response theory, we estimated the discrimination and difficulty parameters for each item and latent variable θ for the political knowledge of respondents. As shown in Table 1, although there was some variation, the discrimination parameter for each item was ensured (Cronbach's α = 0.67). For difficulty, there was substantial variance in the difficulty parameters – indicating some items were considerably difficult, while others were relatively easy. Accordingly, the 10 items used here were considered as a valid measure of political knowledge.

Table 1. Results of applying item response theory to political knowledge items

Table 2 presents the means and standard deviations of the variables of interest of this study. During the recruitment of respondents, an attempt was made to match the age and gender proportion as closely as possible to those of the national census. For the ideology measure, the midpoint of the scale was 6. Although the mean of 6.566 of this measure was slightly inclined toward the conservative ideology, it was statistically significantly lower than the midpoint owing to the large sample size (t[2056] = 11.524); however, the effect size was small (Cohen's d = 0.254). More importantly, this was comparable to the result of a social survey based on national random sampling in Japan. For example, The Utokyo-Asahi Survey (UTAS) (2017) conducted by Masaki Taniguchi of the Graduate Schools for Law and Politics, the University of Tokyo and the Asahi Shimbun data obtained in October 2017 had a mean of 6.206 and a standard deviation of 1.568, while the Comparative Study of Electoral Systems (CSES), module 5 data obtained in February 2018 had a mean of 6.348 and a standard deviation of 1.766.Footnote ⁷

Table 2. Number of valid responses, average values, standard deviation, and correlation matrix for each variable

Note. **P < 0.01, *P < 0.05.

Therefore, it can be said that the composition of respondents in this study does not systematically deviate from the population at large in terms of political ideology.

Table 2 shows the correlation matrix of the variables of interest. Listwise deletion was applied when there were missing values. Militarism and ideology were positively correlated with resistance to apologies. Political knowledge was negatively correlated with resistance to apologies. However, the negative correlation was not so strong that it would pose a serious problem when used as a moderator variable.

To examine whether the association between ideology and resistance to intergroup apology varies by political knowledge, we conducted a multiple regression analysis with ordinary least squares (OLS) estimation (shown in Table 3). In addition to the SDO used for comparison with the ideology measured by respondents' self-reports, demographic variables and militarism, which represents differences in basic values about international relations, were entered in the analysis as control variables.Footnote ⁸ To facilitate the interpretation of the coefficients of the main effects of the variables, conservative ideology and political knowledge were mean-centered. The result revealed that resistance to apologies was significantly associated with SDO, militarism, and conservative ideology. More importantly, the ideology × political knowledge interaction was significant (b = 0.033, S.E. = 0.015, P = 0.026). As Figure 1 shows, the association between ideology and resistance to group apologies was significant only among individuals high in political knowledge, by a simple slope analysis (b = 0.058, S.E. = 0.018, P < 0.001), whereas the association was not significant among individuals low in political knowledge (b = 0.01, S.E. = 0.017, ns). In addition, Figure 2 illustrates the marginal effects of conservative ideology on resistance to apologies at different levels of political knowledge. Figure 2 shows that the association between political conservatism and resistance to group apologies becomes significant around the mean level of political knowledge (as this variable is mean-centered, the political knowledge of 0 in Figure 2 corresponds to its sample mean). Therefore, for individuals with average or higher levels of political knowledge, political conservatism significantly predicted their resistance to group apologies, while for individuals with lower-than-average levels of political knowledge, political conservatism was not significantly associated with their resistance to group apologies.Footnote ⁹

Figure 1. Interaction effect between conservative ideology and political knowledge. Predicted resistance to governmental apologies as a function of conservatism at low (−1 S.D.) and high (+1 S.D.) levels of political knowledge. Prediction is generated from the model shown in Table 3. The values of all other variables are held constant at their respective means.

Figure 2. Changes in the marginal effects of political conservatism on resistance to apologies by political knowledge. Shaded bands represent 95% confidence intervals. Marginal effects and confidence intervals are generated from the model shown in Table 3.

Table 3. Moderation effect of political knowledge on the association between conservatism and resistance to group apologies

Note. OLS model. Conservatism and political knowledge are centralized.

The strong resistance to apologies among people of low political knowledge, regardless of ideology, deserves some speculation. People with low political knowledge may not consider apologies based on the historical and political background under which each of the apologies was made, but rather may make decisions based on their feelings toward the target country. In the results of a poll conducted by Nikkei on 21 January 2019, the percentages of Japanese voters who had negative impressions about China and South Korea were 72 and 61%, respectively.Footnote ¹⁰ Thus, a majority of the Japanese voters today seem to have some negative impression about at least one of the two countries (i.e., the recipients of the governmental apologies that this study utilized). Accordingly, we suspect that respondents with low levels of political knowledge might have responded negatively to the governmental apologies based on their sentiments toward the countries.

Contrary to the analysis with ideology as an explanatory variable, the moderation effect of political knowledge was not significant for the association between SDO and resistance to group apologies. These results are shown in Table 4.

Table 4. Moderation effect of political knowledge on the association between SDO and resistance to group apologies

Note. OLS model. SDO and political knowledge were mean-centered.

To confirm the generality of the aforementioned moderation effect of political knowledge across the three types of past misdeeds, we conducted the same multiple regression analyses for separate issues (i.e., colonial rule, comfort women, and the massacre of Korean people owing to false rumors at the time of the 1923 Kanto Earthquake). The results of a set of three separate multiple regression analyses showed the generality of the results reported in Table 3. Figure 3 shows the regression coefficients and their 95% confidence intervals. The 95% confidence intervals of the regression coefficients overlapped each other for all dependent variables regardless of the issue. Most importantly, the ideology × political knowledge interaction effects were almost equivalent across the three issues.

Figure 3. Determinant factors in resistance to apologies in connection with the three items. They are colonial rule, the comfort women issue, and the massacre caused by false information. Lines represent 95% confidence intervals of coefficients.

4. Study 2

We conducted a replication study to test the moderating effect of political knowledge on the association between ideology and resistance to governmental apologies, and also to examine the two factors that may be associated with among people without political knowledge. As mentioned in Section 2, the first factor is the general attitude of individuals toward apologies, and the second factor is their attitude toward Korea and China, which were victimized by Japan's colonialism policies before and during the war.

4.2. Results

We first examined the reliability of the scale items used in this study. Cronbach's α coefficient for resistance to the three group apologies was calculated to be 0.87.Footnote ¹³ Cronbach's α coefficient of the 16 items measuring SDO was 0.86. The reliability of these scales was satisfactory, as in study 1. The reliability of the Japanese Proclivity to Apologize Measure, a newly added measure in study 2, was also satisfactory at α = 0.89. For political knowledge, the scale was constructed using item response theory, as in study 1. As shown in Table 5, although there were slightly more items with low difficulty than in study 1, a certain degree of discrimination was observed, and the dispersion of difficulty levels was sufficiently ensured. Therefore, it can be said that the 10 items in this study also offered an appropriate scale on which to measure political knowledge.Footnote ^14, Footnote ¹⁵

Table 5. Results of applying item response theory to political knowledge items (study 2)

As in study 1, we conducted a multiple regression analysis using OLS estimation to determine if there was a moderation effect on the association between ideology and resistance to governmental apology (Table 6). In this analysis, we entered demographic variables, SDO, and militarism as control variables. To facilitate the interpretation of the coefficients of the main effects of the variables, conservative ideology and political knowledge were mean-centered. As a result, the regression coefficient of the interaction between ideology and political knowledge was almost the same as in study 1 (b = 0.033, S.E. = 0.018, P = 0.062), although not significant at the 5% level because of the large standard error compared with study 1. As Figure 4 shows, the results of a simple slope analysis showed that the association between ideology and resistance to group apology was stronger for individuals with high political knowledge (b = 0.095, S.E. = 0.018, P ≤ 0.001), than for those with low political knowledge (b = 0.045, S.E. = 0.019, P = 0.018). In study 1, people less informed of political knowledge were more resistant to the government's apology, while in study 2, people with political knowledge were more resistant to the apology. One possible reason for this discrepancy is the salience of the Japan–South Korea conflicts in ordinary Japanese citizens. A few months before study 1 was conducted, a South Korean Navy destroyer allegedly irradiated a Japanese Maritime Self-Defense Force patrol plane with radar. The media coverage of this incident increased the salience of the issue of the Japan–South Korea conflict. As a result, attitudes toward Korea among the politically uninformed, regardless of ideology, became negative, and resistance to an apology might have also been intensified. By contrast, during the period when study 2 was conducted, topics such as COVID-19 and the Tokyo Olympics were widely covered in the press, and the salience of the topic of Japan–Korea and Japan–China conflicts was low. Therefore, only those who had political knowledge and perceived the issue based on conservative ideology responded more negatively to the Japanese government's apology scenarios. In addition, Figure 5 illustrates the marginal effects of conservative ideology on resistance to apologies at different levels of political knowledge. Figure 5 shows that the association between political conservatism and resistance to group apologies becomes significant around the mean −0.7 S.D. level of political knowledge.

Figure 4. Interaction effect between conservative ideology and political knowledge. Predicted resistance to governmental apologies as a function of conservatism at low (−1 S.D.) and high (+1 S.D.) levels of political knowledge. Prediction is generated from the model shown in Table 6. The values of all other variables are held constant at their respective means.

Figure 5. Changes in the marginal effects of political conservatism on resistance to apologies by political knowledge. Shaded bands represent 95% confidence intervals. Marginal effects and confidence intervals are generated from the model shown in Table 6.

Table 6. Moderation effect of political knowledge on the association between conservatism and resistance to group apologies (study 2)

Note. OLS model. Conservatism and political knowledge are centralized.

In addition to the above analysis, we conducted a multiple regression analysis in which we added attitude toward interpersonal apology as an explanatory variable as a factor predicting resistance to group apology among those with low political knowledge (Table 7). The results showed that the main effect of interpersonal attitude toward apology and the interaction effect with political knowledge were not significant at the 5% level.

Table 7. Moderation effect of political knowledge on the association between attitude toward interpersonal apology and resistance to group apologies

Note. OLS model. Attitude toward interpersonal apology and political knowledge are centralized.

Next, we conducted a multiple regression analysis, adding as an explanatory variable the attitudes toward neighboring countries that were affected by Japan's colonial policies in the past (Table 8). The results revealed that the attitudes toward neighboring countries were associated with resistance to apologies. Although the interaction effect between attitudes toward neighboring countries and political knowledge was significant at the 1% level, a simple slope analysis revealed that the association between attitudes toward neighbors and resistance to apology was found among both those with high (b = −0.015, S.E. = 0.002, P ≤ 0.001) and low (b = −0.009, S.E. = 0.014, P ≤ 0.001) levels of political knowledge.Footnote ¹⁶

Table 8. Moderation effect of political knowledge on the association between attitude toward neighboring countries and resistance to group apologies

Note. OLS model. Attitude toward neighboring countries and political knowledge are centralized.

6. Limitations and future research

It should be mentioned that the measurement items used in this study have not yet been sufficiently validated, which is a limitation of this study. For example, resistance to apology was measured using a four-point scale, but the fourth option, ‘do not support at all,’ may have been too strong compared with the other options. Therefore, it may not be appropriate to assume that this item was measured by an interval scale.Footnote ¹⁸ Furthermore, this study relies on a single-item measure that locates each participant's ideological position on a single axis in order to explain the discrepancy in findings in previous psychological research dealing with resistance to group apology. However, given the difficulty for many people to identify their own ideological position on the single axis, it seems informative to operationalize political ideology as a latent variable extracted from responses to divergent items, and examine its relationships with resistance to collective apology.

Among the components of political sophistication, this study used quiz-style political knowledge, which measures the number and extent of cognitive components in voters' belief systems, as a moderator. However, no single method can measure all aspects of political sophistication (i.e., whether voters, like political elites, are able to relate to and understand diverse political events based on an abstract framework such as ideology). Therefore, in the future, it is necessary to examine whether the moderation effects shown in this study can be replicated when political sophistication is measured by the strength of the interrelationships among cognitive factors or by directly measuring understanding of ideological concepts using open-ended responses as in Converse (Reference Converse and Apter1964).

Because the group apology targeted in this study was made in response to complex phenomena such as conflict and historical trauma, there is a limit to examining the determinants of attitudes toward group apologies in quantitative research alone. Therefore, in the future, it is necessary to conduct a qualitative study that examines one case in detail, and to verify the validity of the findings of the current study by comparing them with the results of the qualitative study. In particular, a historical examination of how resistance to apologies by past political leaders affects voters' resistance to apologies by their governments is warranted.

In addition, the most important limitation of this study is that it is not designed to allow inferences about causal relationships among variables. Similar to previous psychological studies that have examined the relationship between ideology and resistance to group apologies, this study employs a correlational research design based on observational data. Moreover, an ‘easy’ linear regression was chosen because there was not enough information to construct a model of how people's resistance to apology is formed. Thus, it is not possible to determine whether conservative ideology in fact foster resistance to collective apology, or their association was merely a spurious correlation due to an unknown third variable. To resolve this issue, experimental studies that test the impact on resistance to collective apology by involving manipulations that promote ideology-based policy understandings or further research using statistical modeling on the relationship between ideology and resistance to collective apology are needed. Previous psychological research examining the relationship between ideology and resistance to apology has assumed that voters are aware of their own political positions on an ideological axis and then link them to policy issues. However, based on the findings of political science on ideology, in this study, we found that the relationships differed based on the level of political knowledge. While governmental apologies are important in encouraging national reconciliation, not all citizens share the same views on apology. One must consider the differences in perspectives of people with different levels of political knowledge while communicating messages that would make it easier for citizens to understand the need for an apology. Thus, despite these limitations, this study's finding that political knowledge moderates the relationship between ideology and resistance to apologies has implications for the study of intergroup apology and policies to promote reconciliation among nations.

This study was limited to Japanese voters. In future studies, it may be possible to determine the reason for the inconsistency in the association between ideology and resistance to group apologies found in prior studies by considering intergroup apologies in various countries and by investigating the moderation effect of political knowledge. We believe that this offers the first step toward that goal.