Islam and Politics: A Latent Class Analysis of Indonesian Muslims Based on Political Attitudes and Psychological Determinants

This study explored the diversity of Muslim political attitudes by conducting a latent class analysis in the rarely investigated context of Indonesia—the largest Muslim country in the world. We surveyed a total of 1208 Indonesian Muslim participants from eight out of 33 Indonesian provinces. The latent class analysis revealed that there are six clusters of Muslim Individuals based on their political attitudes: Fundamentalist Muslim, Nationalist Muslim, Apolitical Muslim, Hijrah Muslim, Moderate Muslim, and Progressive Muslim. Moreover, we also found several meaningful differences in psychological correlates (right-wing authoritarianism, social dominance orientation, and need for cognitive closure) across the six clusters. Taken together, this study sheds some light upon the diversity of Muslim political attitudes and the psychological tendencies that correspond with such attitudes. Information Criteria (BIC) based on the scores of twelve political issues. The BIC estimate was used as a criterion to determine the number of clusters that best fit the data. The BIC estimate informs whether a model (when compared to other models) strikes a better balance between model complexity and explanatory power. A lower value for the resulting cluster or model indicates a better fit (Morgan, 2015). We then analyzed the frequency of response for each cluster in the twelve political attitudes by categorizing the samples in three ordinal categories of attitudes based on their Likert response: 1. Negative (Likert point ‘1’ or ‘2’), 2. Neutral/Moderate (Likert point ‘3’, ‘4’, and ‘5’), and 3. Positive (Likert point ‘6’ and ‘7’). was categorized in a 4-point scale (1 = did not finish secondary school, 2 = finished secondary school, 3 = finished diploma or undergraduate degree, and 4 = finished postgraduate and doctoral study); Income was categorized in a 5-point scale (1 = below IDR 1 million, 2 = IDR 1 million – IDR 2 million, 3 = IDR 2 million – IDR 3 million, 4 = IDR 3 million – IDR 5 million, and 5 = above IDR 5 million). Education and income variables were regarded as ordinal measures (instead of nominal).

Several political psychological studies have viewed Muslims through the lens of Western politics (e.g. Ayoob, 2009;Ibrahim, 2006). Muslims were either minorities or immigrants in this context. Muslim communities can't be lumped into a single political group or merely be part of a discrete dichotomy due to the existence of vastly divergent ideologies within Islam (e.g. Liberal or Conservative, Moderate or Extreme). William E. Shepard published a notable study in 1987 proposing a Muslim political attitude typology (Shepard, 1987). He tried to present Islam as a two-dimensional political ideology. His work divides Muslim political ideology into two axes. First, the endorsement of an Islamic totalitarian system guiding all political, economic, and social behavior was defined as "Islamic totalism. " The second, Islamic modernity, is defined as the acceptance of new technologies, social systems, and the idea of progress. While such research has provided some insight into the diversity of Muslim political views, more research is needed to determine whether Muslims can be mapped in this way. So, the first agenda is to explore the political diversity of Muslim communities.
The dimensional structure is assumed to be inadequate to map Muslim political diversity. Apart from explaining variations in Muslim diversity based on religious interpretations, it is also necessary to map this diversity based on attitudes towards specific political issues. In this study, the author argues that a more detailed typology of Muslim political ideology in Indonesia is needed based on attitudes toward political issues. For these reasons, this study seeks to empirically investigate whether Muslim communities can be classified into more than one political group based on their beliefs, as well as their interpretations of Islamic doctrine and psychological determinants. We surveyed Muslims in Indonesia, the world's largest Muslim country.

Heterogeneity of Political Attitudes in Indonesian Muslim Communities
Since much of the existing literature focuses on the liberal-conservative spectrum (Duckitt, 2001;Erikson & Tedin, 2016), it is unclear whether this knowledge structure is consistent across all contexts within ordinary people's minds (Feldman & Johnston, 2014). This single ideological spectrum acknowledges only the opposing political attitudes within the highly polarized politics of the United States. Though some researchers added a moderate ideological position in the middle of the spectrum (see Branine & Glover, 1997;Kiley, 2017), it remains as a single political spectrum.
In reality, political ideology varies by context. For example, Feldman and Johnston (2014) found that both liberals and conservatives support a wide range of economic and social policies. Moreover, the terms liberal and conservative are arbitrary in many places (Alford, Funk, & Hibbing, 2008). Many cultures do not recognize such terms because political polarization is not based on political ideology but on political figures (Slater & Arugay, 2018) or policy issues (Kinder & Kalmoe, 2017). In non-US contexts, it may be more appropriate to investigate policy preferences based on specific controversial issues within the local context rather than a single ideological dimension.
Religion is deeply ingrained in Indonesian culture (Geertz, 1976;Muluk, Hudiyana, & Shadiqi, 2018). 95 percent of Indonesians living in their own country said religion is important in their lives (Pew Research Center, 2008). Thus, Indonesians are used to evaluating many aspects of their lives from a religious perspective (Mujani, 2007), including the social and economic functioning of society. Within the Indonesian context, Muslim communities may differ in their support for various controversial policies. The two largest Muslim organizations in Indonesia, Nahdlatul Ulama and Muhammadiyah, have been regarded as moderate Muslims, whose followers hold traditionalist or nationalist political stances that preserve Indonesian local wisdom (Barton, 2014). Other Muslim communities stress the establishment of Pan-Islamism, which seeks to replace Indonesia's democratic system with sharia laws (Menchik, 2014). On the other hand, there are also liberal Muslim communities known for their support of pluralism, the secular system, and liberal humanist values (Ali, 2005;Nurdin, 2005).
These groups' supporters may support different political issues. Shepard (1987) proposes that Muslim political attitudes vary widely (e.g. the endorsement of Islamic influence in politics, their stance on economic policies, as well as the support of modernism in daily life). They may support or oppose issues such as banning alcohol, closing nightclubs and bars, and establishing LGBT rights (Tomsa, 2019;Wieringa, 2015) They may also support or oppose religious involvement in public spheres such as mandating Muslim presidential candidates, basing the state on Islam or using religion to determine public policy and even the death penalty for religious blasphemers. They may also support or oppose socialist versus capitalist fiscal systems (Hadiz, 2018), such as whether the government should provide affordable needs for all citizens, raise the minimum wage, reduce the wealth gap, limit foreign investment in Indonesia, etc.
A few studies have been conducted to map Indonesian political attitudes. Edward Aspinall and his colleagues examined the attitudes of various Indonesian political parties to identify political polarization. There was polarization of political attitudes only on religion vs. secular state, not on economic policy or socially progressive stance (Aspinall, Fossati, Muhtadi, & Warburton, 2018). Similarly, Saiful Mujani and his colleagues empirically demonstrated polarization of attitudes on several issues, such as support for establishing sharia laws in Indonesia (Liddle & Mujani, 2007), support for democracy (Mujani, 2007), economic policies (Pepinsky, Liddle, & Mujani, 2012), and political tolerance (Mujani, 2019;Pepinsky, Liddle, & Mujani 2012). He coined the term 'democratic Islam' to describe Muslims who oppose changing Indonesia's democratic system and who are more tolerant of other groups (Mujani, 2007). From this perspective, Indonesian Muslims are divided into two groups: Democratic Muslims and Puritan Muslims.
However, due to differing interpretations of Muslim teachings and power positions in their respective political stages, Indonesian Muslims' political attitudes may not be a single spectrum. Some Muslim schools of thought (denomi nations) embrace modernity and the secular establishment, while others reject it, favoring a more totalitarian Islam (Hudiyana, Putra, Ariyanto, Brama, & Muluk, 2019). This spectrum becomes blurred when we consider that economic endorsements such as capitalism or socialism may not be polarized similarly. To investigate the existence of various Muslim political camps, a cluster analysis that can differentiate political attitudes among Indonesian Muslims is more appropriate. On the basis of Islamic teachings and attitudes towards various political issues, we investigated the latent class structure of the Indonesian Muslim typology.

Psychological Correlates and Political Preferences
Social psychologists have identified dispositional factors that influence political attitudes. Several personalities and dispositional traits have been linked to political ideologies or attitudes. Some scholars argue that personal motivation determines political ideology (Jost, 2006;Jost, Glaser, Kruglanski, & Sulloway, 2003;Jost, Nosek, & Gosling, 2008). They proposed that individuals with higher epistemic motivation-the desire to reduce uncertainty-were more likely to hold conservative political views aimed at preserving the community's traditional system. Thus, those who are more puritanical about policies may have higher epistemic motivation.
Epistemic motivation is the desire to reduce the threat, ambiguity, and complexity of the social world. The need for cognitive closure (NFCC) is defined as the urge to immediately formulate and maintain opinions in order to avoid confusion and ambiguity, and desiring a definite answer (Higgins & Kruglanski, 1996). A person with a high NFCC has closed-mindedness, avoidance of uncertainty, and intolerance for ambiguity. In spite of being exposed to information that contradicts their prior beliefs, people with a high need for cognitive closure tend to choose definitive answers. Conversely, people with low NFCC tend to draw conclusions by gathering and analyzing new information (Brandt & Reyna, 2010).
Social psychologists have found that right-wing authoritarianism or RWA (Altemeyer & Hunsberger, 1992;Duckitt & Sibley, 2010) and social dominance orientation or SDO (Pratto et al., 1994) predict political ideology or orientation. People with higher RWA scores see the world as more dangerous, while people with higher SDO scores see it as more competitive. RWA promotes traditional and religious values such as adherence to the existing social order and respect for authority and tradition. Meanwhile, individuals with high SDO have political attitudes that justify one group dominating another in society, while group hierarchies are seen as fair.
Previous research has suggested that politics and religion are interchangeable because they stem from the same psychological needs (Friesen, Campbell, & Kay, 2015). (Hennes et al., 2012). Thus, religion as a supreme way of life may be correlated with Muslim political attitudes. Religious teachings are seen as absolute morality and should be manifested in society, leading to political attitudes that conform to literal religious teachings. Previous research has shown that higher intratextual religious fundamentalism could explain the perception of threat (Williamson & Hood, 2014), and thus predicted political attitudes or actions (Beller & Kröger, 2018).
We argue that clusters that favor fundamentalism over secularism tend to be higher in various dimensions SDO, RWA, and NFCC. The rationale behind this assumption is that those with higher RWA and NFCC tend to preserve group purity and avoid threats from other groups (Haidt, 2012;Koleva, Graham, Iyer, Ditto, & Haidt, 2012), as well as desire a more simplistic or orderly understanding of the world (Jost et al., 2007;Muluk & Sumaktoyo, 2010). Those who are higher in SDO tend to support the involvement of Islam in public decision making would establish the dominion of Islam in the society. A possible explanation is that these apolitical individuals tend to justify the system such that they may not wish to pursue a political agenda (Meleady & Vermue, 2019;Reese, Proch, & Cohrs, 2014;Stewart & Tran, 2018).

The Present Study
Based on Islamic teachings and political attitudes, this study attempts to determine whether Indonesian Muslim com munities are a single cluster, bi-cluster, or multi-cluster. We didn't predict a certain number of clusters. This could harm the many unexplored possibilities. After identifying the clusters, we analyzed the RWA, SDO, and NFCC scores, as well as demographic data for each cluster.

Method Participants and Procedure
A total of 1208 participants completed our paper-and-pencil survey (N female = 620, 51.32% with M age = 34.20, SD = 11.71). At the time of data collection, 62.83% were married while 31.29% were single, and the rest of the participants (5.88%) had been divorced. A total of 62.58% participants were currently employed, while the remaining 37.42% were unemployed. A total of 77.90% participants finished secondary school as their highest educational attainment, while the remaining 22.10% had completed higher education.
A total of 12.50% participants reported monthly incomes below IDR 1 million (about USD 70). 12.33 percent reported monthly household income between IDR 1 million and 2 million, and 26.16 percent reported monthly income between IDR 2 million and 3 million. Meanwhile, 41.31 percent of participants reported monthly incomes between IDR 3 and 4 million. The rest of the participants (7.70%) reported monthly incomes exceeding IDR 5 million. In 2020, the Indonesian Bureau of Statistics (Badan Pusat Statistik, 2020) reported that the poverty threshold was IDR 2 million per month per household. Thus, 24.83 percent of participants were poor. This number may not represent each province equally and thus should not be used as a benchmark for our study's representativeness. Jakarta's minimum wage is IDR 4.4 million, while West Sumatera's is IDR 2.4 million (WageIndicator.org, 2021).
This study's participants are all Muslims. Participants came from eight Indonesian provinces, each with its own Islamic values, views, and histories (Sakai & Fauzia, 2014). Indonesia has 33 provinces. Jakarta, North Sumatera, and East Kalimantan are considered melting pots with high levels of multiculturalism. West Sumatera and East Java are two other provinces that have shaped Indonesian Islamic traditions. The remaining three provinces are Aceh, Banten, and South Sulawesi, all of which have strong Islamic traditions. We surveyed the participants using stratified sampling, which assigned a city from each of the eight provinces at random. Then we went to the cities with our questionnaires. Participants were given questionnaires after signing consent forms. Participants were debriefed after completing the survey.

Measures
Political attitudes were measured by asking participants the extent to which they agreed with specific political issues. Responses were recorded using a 7-point Likert scale (1 = 'Strongly Disagree', 7 = 'Strongly Agree') for each political issue. A total of twelve political issues were presented: 1. The government should ban alcoholic beverages; 2. The government should close the nightclubs and night bars; 3. The government should not support LGBT rights; 4. This country's president should be Muslim; 5. The system and constitution of this country should be based on religion; 6. Religion should be the primary consideration when making important public policies; 7. Religious blasphemers should be sentenced to death; 8. The government should provide affordable basic goods to the citizens; 9. The minimum labor wage should be increased; 10. The government should regulate the gap between the rich and the poor; 11. The government should regulate economic competition by limiting access of foreign investments; 12. The government should subsidize free electricity, gas, and fuel to the citizens. All other measures (RWA, SDO, NFCC) were adapted to Bahasa Indonesia using back-to-back translation in order to fulfill the cross-cultural adaptation requirement (Beaton, Bombardier, Guillemin, & Ferraz, 2000).
Islamic interpretation was measured by the Totalism in Islam scale. Totalism in Islam was operationally defined as the support for Islam as a totalitarian way of life in all aspects of society (Shepard, 1987). We specifically created 4-items to measure this variable: (1) "Muslims should base their actions by adhering to the Qur'an and Sharia"; (2) "The current solution to the world's problems lies in the essential principles of the Qur'an and Hadits"; (3) "The duty of Muslims is to fight for a society with an absolute rule by Islam"; and (4) "The fate of humankind has been determined, so Muslims should believe that the laws of the Qur'an and Sunnah are final and unchangeable". Participants responded to a Likert Scale of 1-7 (1 = 'Strongly Disagree', 7 = 'Strongly Agree'). The internal consistency score for this measure was α = 0.83. RWA was measured using the adapted version of the 12-item Right-Wing Authoritarianism scale (Altemeyer, 1996), which was re-validated by Passini (2008). There are three dimensions of RWA: submission, aggression, and conventionalism. Participants responded to a Likert Scale of 1-7 (1 = 'Strongly Disagree', 7 = 'Strongly Agree'). The internal consistency score for the submission dimension was α = 0.91, α = 0.76 for the aggression dimension, and α = 0.81 for the conventionalism dimension.
SDO was measured using the adapted version of the 16-item Social Dominance Orientation scale (Ho et al., 2012). The measure consisted of two dimensions, SDO-D (Dominance) and SDO-E (Egalitarian). Participants responded to a Likert Scale of 1-7 (1 = 'Strongly Disagree', 7 = 'Strongly Agree'). The internal consistency score for SDO-D was α = 0.66, and α = 0.65 for SDO-E. NFCC was measured using the adapted version of the Need for Closure scale (Roets & Van Hiel, 2011). Participants responded to a Likert Scale of 1-7 (1 = 'Strongly Disagree', 7 = 'Strongly Agree'). The internal consistency score for this measure was α = 0.72.

Data Analysis
There were no missing data for all of the variables. We did not conduct an exploratory factor analysis (EFA) prior to the main analysis because we wanted to map the Muslim diversity according to relevant political attitudes or behavior, not ideology. This allows us to look at behaviors/attitudes without resorting to ideology. This aligns with previous research that looked at specific political attitudes rather than ideology in the Brexit debate (see Lewis & de-Wit, 2019). We also used Latent Class Analysis (LCA) instead of Latent Profile Analysis (LPA) because LCA is considered superior when the indicator variables are assumed to be independent (Williams & Kibowski, 2015). Because we didn't use EFA to find the latent construct, we treated each indicator variable as a standalone. We did attempt to conduct LPA and found seven classes or profiles (see Supplementary Materials D). However, following Occam's razor (simplicity leads to better data description), we chose the model that was easier to interpret and more parsimonious (Domingos, 1999). The seven-class LPA model is less conclusive and interpretable than the six-class LCA model (see the Results section). So we don't use LPA in our subsequent analyses.
We conducted a k-means solution of latent class analysis (Magidson & Vermunt, 2002;Morgan, 2015) with an estimate from the Bayesian Information Criteria (BIC) based on the scores of twelve political issues. The BIC estimate was used as a criterion to determine the number of clusters that best fit the data. The BIC estimate informs whether a model (when compared to other models) strikes a better balance between model complexity and explanatory power. A lower value for the resulting cluster or model indicates a better fit (Morgan, 2015). We then analyzed the frequency of response for each cluster in the twelve political attitudes by categorizing the samples in three ordinal categories of attitudes based on their Likert response: 1. Negative (Likert point '1' or '2'), 2. Neutral/Moderate (Likert point '3', '4', and '5'), and 3. Positive (Likert point '6' and '7').
Finally, we analyzed the difference between the classes based on their mean scores of RWA, SDO, NFCC, Totalism in Islam, as well as age, participants' income, and participants' education by conducting analysis of variance (ANOVA). We used ANOVA instead of its multivariate counterpart (MANOVA) for two reasons. First, as suggested by Huberty and Morris (1989), the use of MANOVA to control Type 1 error probability is simply illusory. Second, several literatures oppose the practice of using MANOVA because it has to satisfy several assumptions such as observation independence and homogeneity of covariances (Huang, 2020;Pituch & Stevens, 2015). In the context of MANOVA, violation of observation independence best gained through random sampling is very serious (Pituch & Stevens, 2015). As we did not employ a random sampling method, we chose ANOVA instead. Table 1 shows the descriptive statistics for the variables in this study. The best fit latent class solution for political attitudes was a 6-class solution (BIC = 17380.19). The BIC score for 6 class solutions was the lowest. So we concluded that Muslim participants fell into six political clusters. Table 2 and Figure 1 show the complete model fit statistics. In addition, the Lo, Mendell, and Rubin (LMR) test was also conducted and is presented in Supplementary Materials B. From the table, we can infer that the p-value of each class compared to its likelihood was statistically significant in most of the classes. This shows that each model from various classes was acceptable, with the exception of class 18. Contrary to BIC, however, the 7 class solution was more acceptable since it achieved the lowest value. However, consistent with the recommendation from previous work (see Nylund, Asparouhov, & Muthén, 2007), LMR may be less parsimonious as it tends to produce more false-positive results in comparison to BIC. Therefore, we used BIC as our primary estimate.  Class 1 consisted of people who largely rejected Western influences in their personal lives (e.g. alcohol prohibition, nightclub closures, and LGBT rights), largely endorsed Islamic influence in politics (e.g. a Muslim President, sharia laws in the constitution and public policy, and the death penalty for blasphemers), and mostly support the socialist economic agenda (e.g. rich-poor gap control; the increase in labor minimum wage; subsidy and support of basic needs; limiting the economic competition with foreign investors). We named this class "Fundamentalist Muslim" after Islamist movements' puritanical attitudes. People in Class 2 largely rejected Western influence in their personal lives, endorsed the socialist economic agenda, but were moderate in their support for Islamist political influence in the country. We called this group "Nationalist Muslims" because they favored traditional Indonesian secular political systems over Islamist political systems. Class 3 represented people who were neutral or moderate on most issues (relative to other classes). We named this class "Apolitical Muslim" due to their apparent lack of political conviction. Note. BIC = Bayesian Information Criteria; χ 2 = Chi-square; df = degree of freedom.

Bayesian Information Criteria (BIC) for Each Number of Clusters
As for Class 4, it mainly consisted of individuals who rejected the Western influences toward private life and mostly en dorsed the establishment of Islamist politics but had a tendency to be moderate in their support towards the socialist (vs. capitalist) economic system. We named this class "Hijrah Muslim"-named after a newly observed Muslim community that showed a tendency to endorse the capitalist economic system but reject Western influence and traditionally secular politics. Class 5 consisted of individuals who showed a more balanced response in most issues, with the exception of issues that are considered normative for Indonesian people (e.g. rejection of LGBT rights; socialist economics). Since these individuals showed a tendency to exhibit more normative attitudes in Indonesian public norms (relative to other classes), we named this class "Moderate Muslim". Finally, Class 6 included individuals who were less likely to reject Western influence in private life, less likely to endorse Islamist politics (compared to any of the other classes), and mostly endorse the socialist economic system. These attitudes may reflect a more humanist and modern stance, and thus we named this class "Progressive Muslim".
We then examined whether the scores of Totalism in Islam differed between classes by conducting analysis of variance (ANOVA). Class 1 (Fundamentalist Muslim) was found to possess the highest mean score for Totalism in Islam, while Class 5 (Moderate Muslim) was found to possess the lowest mean score. We found that the six classes indeed possessed different degrees of Totalism in Islam. Next, we examined whether individuals across different classes exhibited different degrees of psychological corre lates. Table 3 illustrates the mean scores, standard deviations, as well as analysis of variance results for all psychological correlates across the six classes. Post-hoc analyses for all psychological correlates are illustrated in Supplementary Materials C. We found that Class 1 scored higher in NFCC compared to any of the other classes, while there were no observed differences between the remaining classes. This means that those in the Fundamentalist Islam class were more likely to be oriented toward cognitive certainty and order than any of the other classes. Note. N = number of samples in a class; M = mean; SD = standard deviation; ANOVA = analysis of variance; F = F-test for mean difference; df = degree of freedom; p = p-values; RWA = right-wing authoritarianism; SDO-D = social dominance orientation -dominance; SDO-E = social dominance orientation -egalitarian; NFCC = need for cognitive closure; Education was categorized in a 4-point scale (1 = did not finish secondary school, 2 = finished secondary school, 3 = finished diploma or undergraduate degree, and 4 = finished postgraduate and doctoral study); Income was categorized in a 5-point scale (1 = below IDR 1 million, 2 = IDR 1 million -IDR 2 million, 3 = IDR 2 million -IDR 3 million, 4 = IDR 3 million -IDR 5 million, and 5 = above IDR 5 million). Education and income variables were regarded as ordinal measures (instead of nominal).
For the submission dimension of RWA, Fundamentalist Muslims scored higher when compared to Apolitical and Hijrah Muslims, but didn't score higher in comparison to Nationalist, Moderate, and Progressive Muslims. Nationalist, Moderate, and Progressive Muslims were all found to score higher when compared to Apolitical and Hijrah Muslims. Thus, both Apolitical and Hjrah Muslims were observed to be the lowest in scores when we compared them to the other classes in the submission dimension. As for the aggression dimension of RWA, Fundamentalist Muslims scored higher in comparison to any of the other classes, while Progressive Muslims scored higher only when compared to Nationalist Muslims and Hijrah Muslims. There were no other differences observed for the aggression dimension. Finally, for the conventionalism dimension of RWA, Fundamentalist Muslims scored higher in comparison to Nationalist and Progressive Muslims. Moreover, Nationalist Muslims scored higher in comparison to Apolitical, Hijrah, and Progressive Muslims. These findings suggest that Fundamentalist Muslims were more likely to be oriented toward right-wing aggression. However, Fundamentalist Muslims did not differ in right-wing conventionalism and submission when we compared them to Moderate Muslims. Fundamentalist Muslims scored higher, however, in comparison to Nationalist Muslims and Progressive Muslims in right-wing conventionalism. Fundamentalist Muslims also scored higher when compared to Apolitical Muslims and Hijrah Muslims in right-wing submission. Meanwhile, for SDO-D or the dominance dimension of SDO, Fundamentalist Muslims scored higher compared to all other classes, with the exception of Apolitical Muslims. Moreover, Apolitical Muslims scored higher when compared to Nationalist, Moderate, Hijrah, and Progressive Muslims. Thus, Fundamentalist and Apolitical Muslims ranked as the highest and the second-highest in the SDO-D dimension respectively. As for SDO-E or the egalitarian dimension, we found a similar pattern as with SDO-D, in which Fundamentalist Muslims scored higher compared to any of the other classes, with the exception of Apolitical Muslims. Though Apolitical Muslims scored higher in SDO-E compared to Nationalist, Moderate, and Hijrah Muslims, they did not score higher compared to Progressive Muslims. This pattern is intriguing because evidently, anti-egalitarian orientation was lower for Nationalist Muslims, Moderate Muslims, and even Hijrah Muslims, in comparison to Progressive Muslims.

Figure 3
The Distribution of Gender for All Classes

Discussion
The latent class analysis of 1208 Indonesian Muslims revealed that Muslims cannot be simply dichotomized. Muslims' political views are more varied. The Bayesian Information Criterion best fit six Indonesian Muslim clusters. We named each cluster: Fundamentalist (Class 1), Nationalist (Class 2), Apolitical (Class 3), Hijrah (Class 4) and Moderate Muslim (Class 5), and Progressive Muslims (Class 6). We looked at the RWA, SDO, and NFCC mean scores for each class and found some interesting patterns. Demographic variables (age, income, education) did not differ for each class except for gender.
Individuals in the Fundamentalist Muslim cluster (Class 1) endorsed the prohibition of alcoholic beverages, the denial of LGBT rights, and the closure of nightlife. These attitudes reflect ultra-conservative tendencies that seek to protect the group's purity from outside threats (Haidt, 2012;Koleva et al., 2012;Skitka & Tetlock, 1993). They also endorsed the establishment of sharia laws in the constitution and public policy and elected a Muslim figure as president, as was the punishment of blasphemers. They consistently outperformed all other classes on the Totalism in Islam scale. The rise of religious fundamentalism, which opposes secularization, may explain their popularity (Stark, 1999;Zubaida, 2005). Fundamentalist Muslims also scored higher on the NFCC, the RWA aggression dimension, and all SDO dimensions. They also tend to have more simplistic cognition (Muluk & Sumaktoyo, 2010), a stronger need for certainty and order (Jost et al., 2007), hostility towards weaker minority groups (Duckitt, 2006), and a higher tendency to endorse intergroup hostility (Duckitt, 2006). (De Zavala, Cislak, & Wesolowska, 2010).
They (Class 1) also strongly oppose the capitalist economic system. This finding appears to contradict prior research showing fundamentalist groups were motivated by symbolic rather than economic threats (Jackson & Esses, 1997;Mashuri et al., 2016). However, we argue that in Indonesian politics, Chinese Christian minorities wield economic power within a capitalist financial system, opposing Muslim majorities who resent the aforementioned minority group. Like Amy Chua's "World on Fire" hypothesis, Southeast Asia's free market democracy may breed intergroup hatred (Chua, 2004). A socialist religion, Islam is arguably a socialist religion with various teachings aimed at establishing economic equality (Hefner, 2017). (Naqvi, 1994(Naqvi, /2013. Those in the Nationalist Muslim (Class 2) and Moderate Muslim (Class 5) clusters tended to oppose Islamization in the public sphere. These two clusters differed in their strong attitudes toward issues of social inclusion. Nationalist Muslims generally take a neutral or normative stance on most issues. They followed more religious doctrines, which mainstream Indonesians value. Conversely, moderate Muslim participants were more likely to support socially inclusive issues. For example, they were more accepting of non-Muslim political leaders, more accepting of LGBT people, and more supportive of secular laws not based on their own religion. These two clusters shared similarity in many psychological traits (RWA, SDO, and NFCC) and Totalism in Islam scores. The Nationalist Muslim cluster had more females than the Moderate Muslim cluster. They also scored lower than the Fundamentalist Islam cluster in RWA, SDO, and NFCC.
The Progressive Muslim cluster (Class 6) was the most moderate stance in anti-Western attitudes (LGBT rights, closing of nightlife activities, and banning of alcohol). They tended to be more socialist on economic issues and less likely to support Islamization of the public sphere. However, compared to Nationalist and Moderate Muslims, Progressive Muslims scored higher on the Totalism in Islam scale. This finding reflects modern Muslims' acceptance of a globalist Muslim identity that seeks social justice for Muslim identities in various public domains (Kundnani, 2008). This is merely a hypothesis that should be investigated further. Compared to other classes, Progressive Muslims scored lower in RWA conventionalism. This may be because progressive liberal communities tend to be more supportive of social change and progress (Haidt, 2012;Osborne, Wootton, & Sibley, 2013). However, compared to Nationalist or Moderate Muslims, these people tend to be more egalitarian. Class 6 has the most egalitarian political attitudes of all the classes, so this result is puzzling. Although they oppose ultra-conservative groups (but not the secular status quo in Indonesia), it is possible that Class 6 has a more non-egalitarian orientation toward conservative groups. These people may be less likely to support fundamentalist social justice visions that seek to replace the secular status quo with Islam. Of course, this is all hypothetical, and future research should provide empirical evidence.
Individuals in the Apolitical Muslim cluster (Class 3) tend to be neutral or moderate on most issues. However, many of these people support the prohibition of alcohol and oppose LGBT rights. Remember that they are more moderate than Fundamentalist, Nationalist, and Moderate Muslims (higher frequency of neutral responses to the two issues). They ranked last in RWA's submission and aggression dimensions, along with Hijrah Muslims (Class 4). Apolitical Muslims may be less likely to be ideological because they are less concerned about all political issues (as evidenced by their mostly neutral attitudes) (Jerolmack & Porpora, 2004;Monroe & Kreidie, 1997). They also ranked second in SDO-D, the SDO dominance dimension. This result matched our explanation that Class 3 (Apolitical Muslim) tends to maintain the status quo in society and thus has less political attitudes.
Finally, the Hijrah Muslim cluster (Class 4) had more moderate support for socialist economic policies, but strong support for anti-Western policies and the establishment of Islamic influence in politics. To promote the Islamic identity, Hijrah communities sell exclusive and luxurious goods to promote the Islamic identity, and promote conservatism through the free market system (Amna, 2019;Hasan, 2019). We couldn't find much literature on this topic. This issue may be unique to Indonesia, where capitalism and Islam coexist (Hadiz, 2011). But, like the Fundamentalist Class, this Hijrah Class may express hatred towards perceived enemies (Mashuri & Zaduqisti, 2019). (Muluk, Sumaktoyo, & Ruth, 2013). The main difference is their acceptance of capitalism.
Across all classes, we found an overwhelming anti-Western influence on personal issues (banning alcohol, rejecting LGBT rights, closing nightclubs and bars). The rise of conservatism in Indonesia over the last decade may be blamed on these findings (Menchik & Trost, 2018;Mietzner & Muhtadi, 2018). Our study participants from the Fundamentalist Islam cluster (Class 1) showed this tendency. There were 462 Fundamentalist Muslims, far more than any other class. Despite the fact that this study showed that Muslims are very diverse, a third of the participants were puritanical and ultra-conservative.
As shown in Figure 2, although they tend to be neutral in attitudes towards supporting Islamic law, progressive Muslims are more supportive than moderate Muslims. This could be explained like this. In Indonesia, moderate Muslims often support traditional Indonesian values in opposition to sharia supporters (Robinson, 2011). On the other hand, progressive Muslims, drawn to Western liberal democracy, may not share traditional Indonesian values. But when it comes to other issues related to the Islamic law establishment, Progressives and Moderates have very similar views.
Other than providing novel insight regarding Muslim political camps, our findings revealed that fundamentalist Muslims who are more supportive of economic equality policies should be distinguished from Hijrah Muslims who welcome the free-market economy-even though both represent the face of religious fundamentalism. In addition, Muslims such as Nationalist and Moderate Muslims should be distinguished from Progressive Muslims. While the three groups seem to agree that secular laws (vs. sharia-based laws) should be maintained, these groups disagree on the issue of attitudes towards Western lifestyles.
This classification may help policymakers predict social movement narratives. For example, the political agenda underlying collective actions may vary between Muslim groups. Fundamentalists may use economic injustice and symbolic threat as narratives for their movement, whereas Hijrah Muslims may not. Meanwhile, progressives may fight for marginalized groups like the LGBT community. Nationalists may oppose this movement because they believe it threatens Indonesia's traditional family values.
In addition, we can map Indonesian political figures and parties based on the categories we established in our research. Such categorization may prove to be more meaningful in comparison to classifications based on a single political spectrum, such as liberal vs. conservative or modern vs traditional. With this in mind, we can predict which political candidates or parties will acquire support from each political camp.
We noted several limitations to this study. First, the latent class samples may not be sufficiently representative, as we only collected data from eight of Indonesia's 33 provinces. But we chose the eight provinces carefully, as they represent Indonesia's various Islamic traditions. Second, our sampling method was more convenience-based than systematic random sampling. This condition may jeopardize external validity. Third, we didn't cover other relevant political issues like women's rights, freedom of expression, and solidarity with other Muslims (e.g., Uighur Muslims in China, Rohingya Muslims in Myanmar, Palestine Muslims). Inclusion of such issues may provide a broader picture of Muslim diversity. Future research should replicate this study with larger samples using a probabilistic method from all Indonesian provinces and focus on political issues. Fourth, we did not consider the correlates like intolerance and intergroup hostility. Future research should consider these variables to better understand and predict potential problematic groups that could cause conflict among all clusters.

Conclusion
We found six clusters of Muslim individuals based on political attitudes through latent class analysis: 1) Fundamentalist Muslim; 2) Nationalist Muslim; 3) Apolitical Muslim; 4) Hijrah Muslim; 5) Moderate Muslim; and 6) Progressive Muslim. These clusters differed in several psychological correlates such as Need for Cognitive Closure (NFCC), various dimensions of Right-Wing Authoritarianism (RWA), and various dimensions of Social Dominance Orientation (SDO). Therefore, we conclude that Muslim individuals cannot be categorized by a single spectrum or a dichotomic differentia tion.