Memory of Conflicts and Perceived Threat as Relevant Mediators of Interreligious Conflicts

Abstract

The present study investigated to what extent memory of conflict and perceived threat explain the relation between religiosity and supporting interreligious conflicts between Muslims and Christians in Indonesia. We employed data from the survey of the interreligious conflicts in 2017, involving 2026 adults from five hotspot regions: Aceh Singkil, South Lampung, Bekasi, Poso, and Kupang. Our confirmatory factor analysis and measurement invariance demonstrated that all employed scales were valid and reliable across religious groups. Our structural equation modelling showed that while the memory of conflicts was only positively related to supporting lawful protests, the perceived threat was shown to be strongly related to supporting both lawful and violent protests. This shows that memory of past physical injuries is not highly susceptible to exclusive behaviours against the religious outgroup. However, it is the individuals’ evaluation of the religious outgroup as a result of past conflicts which encourages exclusionary behaviours against them. These findings provide empirical insights into the importance of the aftermath of interreligious conflicts and how they can be used to avoid future clashes.

1. Introduction

In the past two decades, religious differences have increasingly become a focus in public life domains across the world (e.g., Ben-Nun Bloom et al. 2015; van Bruinessen 2018; Wright 2016). One possible explanation is the awakening of the long history of interreligious conflicts involved in a nation, such as in Myanmar (see Kipgen 2013) and in Indonesia (see Human Rights Watch 2013). This fact is nothing new and has long been predicted by Coser (1956); two religious groups living side by side will naturally have some degree of competition over actual and perceived scarce resources (Scheepers et al. 2002b). This, in turn, intensifies individuals’ religious identification and the possibility of conflicts. In addition to contextual factors like group size, power and status differences between religious groups, historical factors such as past conflicts have also been shown to be related to the possibility of future interreligious conflicts (Schlueter and Scheepers 2010; Bar-tal 2007).

Past conflicts tend to leave memories of traumatic experiences among the involved parties even when the conflict itself has subsided (Bar-tal 2007). These memories are eventually passed on from one generation to the next, maintaining exclusionary measures against the religious outgroup. Extensive sharing of collective memory is enabled through a shared religious identity. This is possible because when individuals identify with a certain religious group, they consciously distinguish themselves from the non-believers in terms of religious beliefs, practices, and experiences (Turner 1975; Tajfel and Turner 1979; Stark and Glock 1968). Furthermore, shared religious identity provides cues for the interpretation and evaluation of given information about the individuals’ ingroup as well as the outgroup (David and Bar-Tal 2009). Therefore, religious ingroup members can easily relate to the collective memory of interreligious conflicts.

In addition, the memory of conflicts has been shown to have the potential to manifest into other interreligious conflicts, especially when the current situation threatens the religious ingroup’s identity (Wohl and Branscombe 2008; Tajfel and Turner 1979). This is due to the memory of conflicts being accompanied by perceptions of threat towards the relevant religious outgroup. Indeed, conflict is both a response to the threat perceived by ingroup members as well as past encounters with the outgroup(s) (Quillian 1995; Beller and Kröger 2017; Schlueter and Scheepers 2010). However, the main reason is that the type of group identity involved in interreligious competition and conflict, namely religious identity, is not merely about religious beliefs. It also refers to one’s social position (Ysseldyk et al. 2010; Setiawan et al. 2020). By identifying with a particular identity, individuals naturally conceive their ingroup vis-à-vis the relevant outgroup(s) in terms of their size, i.e., majority vs. minority, as well as their power, i.e., dominant vs. subordinate (Blumer 1958; Bobo 1999). Therefore, not only is religious identity directly related to religious ingroups’ attitudes towards the outgroup, but it is also related to the level of threat perceived by religious ingroup members.

Thus far, we have argued that memory of conflicts and perceived threat can further explain the relation between religious identity and the possibility of future interreligious conflicts. Yet, we have little empirical insight into the validity of this claim in areas where interreligious conflicts have erupted. To further complicate, religious identity is a complex construct, distinct from other types of social identity. Beliefs held in religious identity are to be taken wholeheartedly and conceived to be the most righteous by the adherents (Ysseldyk et al. 2010; Stark and Glock 1968). This partly explains the emergence of religious fundamentalism or extremism, e.g., ISIS (Herriot 2014; Beller and Kröger 2017). On the other hand, embracing religious identity also brings about positive consequences, such as caring for the poor and loving one’s neighbours (Setiawan et al. 2020; Glas et al. 2018). Therefore, it reinforces the need to investigate the potential difference of play between each religiosity dimension on one hand and memory of conflict and perceived threat on the other, which in turn, impacts individuals’ attitudes towards interreligious conflicts.

To understand this, Indonesia provides a unique case for the current study. Its constitution guarantees freedom for its citizens to practice any of the six official religions. Yet, the country consistently experiences interreligious conflicts, mostly between Muslims (87% of the country’s population) and Christians (11% Protestant and Roman Catholic combined) (Badan Pusat Statistik 2021; Pew Research Center 2015; United States Department of State 2017). In parts of Indonesia where Muslims are the majority, they are often found to perpetrate conflicts (see religious freedom report by Halili (2016) and Human Rights Watch (2013)). Otherwise, in other parts where Christians are the majority, such as Kupang and Papua, they are found to be the offender. This situation reflects the majority–minority position dictated by religious identity. In addition, religion has increasingly taken a more prominent role in public domains in Indonesia. Suggestions like voting only for a regional and national leader from a religious ingroup and even implementing Sharia laws in several parts of the country are only a handful of examples of how religion has become pervasive in individuals’ social life (Mulia 2011; Hadiz 2017).

Taken together, interreligious conflicts are deeply rooted in the way people identify with their religious identity, and are encapsulated in a notion of religiosity (Stark and Glock 1968; Wright and Young 2017; Wibisono et al. 2019; Ysseldyk et al. 2010). To predict future conflicts, however, also requires looking at traces of past conflicts (Bar-tal 2007). Along with the memory of conflicts, perceived threat appears to be a constant byproduct of interreligious tension after conflicts subside (Beller and Kröger 2017; Kanas et al. 2015). Therefore, we expect that religiosity is related to supporting interreligious conflicts among Muslim and Christian communities in Indonesia, and this relation can be explained by memory of conflicts and perceived threat among the religious groups.

To fill in the lacunae in the current literature, this study seeks to investigate to what extent memory of conflicts and perceived threat explain the relation between religiosity and supporting interreligious conflicts in Indonesia. This investigation is built upon previous studies that have paved the way for the explanation of the relation between religious identity and attitudes towards the religious outgroup. Studies of Scheepers and Eisinga (2015) and Setiawan et al. (2020) have shown that the extent of individuals’ religious identification, based on Tajfel (1974), can be explained by assessing the level of religiosity dimensions of Stark and Glock’s (1968). Their studies demonstrate that the most relevant religiosity dimensions considered in explaining exclusionary measures towards the outgroup are religious beliefs, religious practices, and consequential dimension or religious salience.

Specifically, we aim to undergo the following strategies in this study. First, drawn from the previous studies and reports, interreligious conflicts will be studied as latent behaviour consisting of two modes to be able to explain future conflicts: (1) supporting lawful protests and (2) supporting violent protests. Reports have shown that interreligious violence often starts from mere demonstration which then escalates into physical clash (Halili 2016; Human Rights Watch 2013). Therefore, it is vital to take into account both modes of interreligious conflicts to capture the individuals’ attitudes towards public demonstrations at the expense of the religious outgroup and the extent to which individuals are willing to go further beyond demonstrations. Second, using the obtained dataset made publicly available by Setiawan et al. (2018), the proposed study employs samples of ordinary citizens to advance our understanding of interreligious conflicts in Indonesia, as previous studies mostly relied on student populations (e.g., Kanas et al. 2015; Pamungkas 2015; Subagya 2015). Hence, this study provides empirical insights into the matter of interest among the general population in Indonesia. Third, we will perform measurement invariance for all measures employed as a means to acknowledge religious group differences and ensure that all scales are equally understood by both religious groups, thus improving the applicability of all scales in the Indonesian context.

2. Theories and Hypotheses

The study synthesizes the notions of social identity theory (Tajfel 1981; Tajfel and Turner 1979), competition and threat theory (Croucher 2017; Scheepers et al. 2002b), and memory of conflicts (Bar-tal 1998) to formulate and empirically test the model which explains the support for interreligious conflicts. We will start by explaining social identity theory in the scope of religious identity and how this notion can be used to reflect religiosity and one’s social position. Next, we explain how shared religious identity helps develop and cement memory of conflicts and perceived threat among ingroup members. All these notions are then synthesized into a theoretical model that explains the extent to which individuals support interreligious conflicts.

2.1. Religious Identity

Based on Turner’s (1975) work on social identity, individuals innately identify themselves with certain social groups as an attempt to make sense of their social world. Once they define their social identity, they distinguish themselves from others who are not part of the selected identities (Tajfel and Turner 1979). As part of maintaining their social identity, individuals constantly compare their membership to the relevant groups by generally viewing their ingroup favourably while viewing the outgroups less favourably. This tendency has two functions, one is to maintain or achieve superiority and the other is to keep their psychological distinctiveness (Phinney and Ong 2007). This distinctiveness is then transformed into behavioural and affective outcomes towards the outgroups.

Religious identity, however, is not only related to beliefs and rituals. It also brings about one latent consequence which is often found to be the main cause of interreligious competition, the sense of social position. This notion is borrowed from Blumer’s (1958) thesis which proclaims that racial groups divide population into different social positions about proprietary claims, i.e., majority vs. minority. It is worth mentioning that a sense of social position refers to the position of religious ingroup vis-à-vis relevant outgroup(s), not individuals. Thus, somewhat reflecting the context of Indonesia, religious identity is strongly related to one’s social position. Being the most Muslim-populated country, identifying with Islamic religion entitles individuals to a majority position while the rest is a minority (Pew Research Center 2015). This enables cases where the majority is often found to claim privileges over scarce resources while, at the same time, the minority appeals for a better distribution of those resources (Hadiz 2017). Nevertheless, Indonesia adds nuance to social position as related to religious identity. The Muslim majority in Indonesia is also often found to generate narratives of long suppression during the Suharto era in which they claim that they have not received the proprietary claims that they should get and this has enabled the Christian minority to threaten their majority position (van Bruinessen 2018).

As previously mentioned, identifying with a religious group means identifying with all components that come with it; not only a sense of their social position. Individuals who identify themselves with a particular religious identity means adopting the beliefs and rituals of that particular religion (Ysseldyk et al. 2010). This is in line with the notion of religiosity, in which people’s religious identification is defined by looking at different dimensions of religiosity, e.g., practices and beliefs (Stark and Glock 1968; De Jong et al. 1976; Holdcroft 2006; Jankowski et al. 2011). Based on the notion above, we focus on religiosity dimensions that have been shown to relate to interreligious conflicts, namely religious practices, beliefs, and salience (Pieterse et al. 1991; Anthony et al. 2015).

The religiosity dimensions of interest echo social identification theory which involves dimensions of ingroup ties, ingroup affect, and centrality (Cameron 2004). Specifically, the extent to which individuals practice their religious rituals largely reflects their religious ingroup ties. Additionally, the extent to which individuals affectively evaluate their religious beliefs is a reflection of ingroup affect. Finally, the extent to which individuals rely on their religious values on daily life basis strongly reflects the centrality of their religious identity in their life.

Moreover, apart from religiosity dimensions, we are also aware that there are individual characteristics that may partake in determining individuals’ social position in relation to supporting interreligious conflicts. By including these characteristics, we can ensure that there will be no spurious relationships between variables of interest when individual characteristics are factored in. Specifically, we include age, gender, education, and income. Beller and Kröger (2017) demonstrated that women show lower support for extremist intergroup violence. Further Barron et al. (2009) and Humaedi (2014) found that social gaps in education and income play a big role in interreligious conflicts. Subsequently, all three religiosity dimensions are further delineated separately in the following sub-sections.

2.2. Religious Practice

Performing religious practice is assumed to strengthen ingroup ties by increasing coalitional commitment and cooperative behaviours (Ginges et al. 2009). These practices can be reflected in the extent to which individuals perform rites and liturgical acts regularly (e.g., attendance to religious services) (Anthony et al. 2015; Stark and Glock 1968). Through attending to religious services, ingroup members can interact and exchange ingroup norms and values. Combined with the natural tendency of maintaining ingroup psychological distinctiveness, individuals who frequently attend religious services are expected to display more typical ingroup behaviours in times of interreligious tension (Ginges et al. 2009; Tajfel and Turner 1979).

A recent study by Beller and Kröger (2017) shows that there is a positive relation between regular mosque attendance and supporting intergroup extremist violence. On the contrary, another study shows that performing religious practice negatively relates to violent jihad among Indonesian Muslims (Muluk et al. 2013). This contradiction makes religious practices even more interesting to study whether cooperative behaviours from religious practice hold in areas where a large scale of interreligious conflicts have erupted. Given the condition that those who regularly perform religious practices have a higher chance to share religious ingroup experiences, we expect that religious practice is more likely to be related to ingroup collective action. Based on the foregoing, the first hypothesis of the study is formulated as follows:

Hypothesis 1 (H1).

Religious service attendance is positively related to supporting interreligious conflicts.

2.3. Religious Beliefs

As the heart of faith, religious beliefs reflect how individuals feel towards their religious identity (Stark and Glock 1968). This is defined by the extent to which they positively evaluate their religious doctrines, while seldomly negatively evaluating the religious outgroups’ doctrines. In the meantime, religious doctrines are assumed to provide a positive ingroup-affect of individuals’ religious identity and, thus, they strive to maintain the religious ingroup’s psychological distinctiveness (Cameron 2004; Turner 1975). By embracing a particular set of religious beliefs, individuals are prone to particularistic views, in which they tend to see their religion as the only true vision of the supernatural being (Stark and Glock 1968). In addition, individuals’ beliefs are also related to religiocentrism, i.e., the combination of positive attitudes towards one’s religious ingroup and negative attitudes towards religious outgroups (Brewer 1999; Sterkens and Anthony 2008). A recent study on interreligious conflicts in Indonesia shows that particularistic views are related to supporting lawful protest, but not a violent one. However, religiocentrism is found to be related to both types of protest.

Throughout interreligious conflicts, it is expected that ingroup members ingrain themselves even more in their religious community (Tajfel and Turner 1979), not only for personal security but also to maintain the positive aspects that they believe in their religious identity. Therefore, it is common to learn that a religious ingroup develops narratives of conflict justification as well as victimization after the conflict ends, passing them on to later generations (Bar-tal 1998; van Bruinessen 2018). This is one way to maintain their superiority or positive group image among religious ingroup members. Of course, this means that the other side of the conflicting party also does the exact socialization. Hence, not only does collective memory of conflicts persist among ingroup members, but perceptions of threat towards a religious outgroup should also remain due to the religious beliefs-imbued conflict experience. Therefore, the hypotheses regarding religious beliefs are as follows:

Hypothesis 2 (H2).

Particularism is positively related to supporting interreligious conflicts.

Hypothesis 3 (H3).

Religiocentrism is also positively related to supporting interreligious conflicts.

2.4. Religious Salience

Finally, religious salience refers to the centrality of religious values to individuals (Roof and Perkins 1975; Stryker and Serpe 1994). Identifying with a particular religious identity does not merely affiliate oneself with the community, but also with religious values. These values, as previously mentioned, are considered by individuals as the positive aspects that they cling to. Therefore, religious salience goes beyond simply adhering to religious communities, as reflected in ingroup ties and affect. Religious salience is more about being a good and faithful Muslim or Christian (Roof and Perkins 1975). As such, those with a high level of religious salience may tend to shy away from interreligious conflicts because they rely on positive religious values, e.g., love thy neighbours, forgiveness (Glas et al. 2018). This claim is supported by Scheepers et al. (2002a) who found a negative relation between religious salience and outgroup prejudice.

Based on the nature of religious salience, this dimension may also relate to a low tendency of reliving a memory of conflicts among ingroup members. Further, those who place positive religious values in a central position in their life are more likely to perceive outgroup members benignly. This is possible due to their emphasis on positive religious values, such as forgiving others and helping people in need, regardless of their religious communities’ attitudes. Taken together, we hypothesize the relation between religious salience and supporting interreligious conflicts as follows:

Hypothesis 4 (H4).

Religious salience is negatively related to supporting interreligious conflicts.

2.5. Memory of Conflicts

In areas where a large scale of interreligious conflict has once erupted, the relation between religious identity and supporting interreligious conflicts is more likely to be further explained by past encounters between the two conflicting groups. In this study, we propose two mediators be investigated, namely memory of conflicts and perceived threat. We start by explaining the memory of conflicts.

The basic component of the memory of conflicts is societal beliefs (Bar-tal 2007). These beliefs are cognitions shared by group members on topics and issues that are important for their group and to their distinctiveness. Regarding interreligious conflicts, each religious group may have certain societal beliefs that influence the course of conflicts, such as the own group’s justification of conflicts, degrading the outgroup, victimization, patriotism, unity, and peace (Bar-tal 1998). By holding certain societal beliefs, ingroup members are given social cues for the interpretation and evaluation of given information. As such, a memory of interreligious conflicts develops over time and the societal beliefs lay out these memories in a coherent and meaningful piece (Bar-tal 2007). This is also largely propped by the characteristics of collective memory. First, these memories do not intend to give clear and objective history, but rather a functional history to unite the members in times of conflict. Therefore, a memory of interreligious conflict is likely to contain false and biased information in ways that fulfil society’s needs at that time. Second, collective memory is treated as the true history of society. This memory is often displayed in school textbooks. Further, this memory is firmly established in a specific socio-political-cultural context through formal and informal socialization among ingroup members (Bar-Tal et al. 2009).

Based on the notions above, a memory of conflicts takes an important part in escalating or invoking interreligious conflicts for several reasons (Bar-tal 1998). First, it justifies a religious ingroup’s involvement in conflict (Bar-tal 2007). Second, it provides a positive self-image of the ingroup. Feelings of patriotism rise when ingroup members take part in defending their religious group. Third, it delegitimizes the opponent (Bar-tal 1998). Fourth, it paves the way to a belief in victimhood. In the Indonesian context, the narrative shared among the Muslims is the repression of Islamic ideas by Westernization (van Bruinessen 2018). At the same time, Christians feel that they have long been oppressed, especially in terms of religious practices (Human Rights Watch 2013). With each religious group believing that they have endured long suffering and injustice, this makes the peace-making process an almost impossible phase to reach. However, measuring the collective memory of interreligious conflicts in Indonesia can be complicated. One, interreligious conflicts are not included in a school textbook. Two, discussions related to religious differences, let alone interreligious conflicts are mostly banned from public discussion (Human Rights Watch 2013). Therefore, we opted to capture a memory of conflicts based on the frequency of talks about past interreligious conflicts. The hypotheses related to the memory of conflicts are formulated as follows:

Hypothesis 5 (H5).

Memory of conflicts positively mediates the relations between religious practice and beliefs, on one hand, and supports interreligious conflicts on the other.

Hypothesis 6 (H6).

On the contrary, a memory of conflicts negatively mediates the relation between religious salience and supporting interreligious conflicts.

2.6. Perceived Threat

Perceived threat can be defined as a sense of awareness of the challenge brought by outgroups, i.e., minority (Blalock 1967; Olzak 2013). The reason is twofold. Perceived threat is a subjective evaluation of individuals towards their competition with the religious outgroup (Blalock 1967). Even when the actual competition is low, identified by religious composition and economic division among the Muslims and Christians in certain areas, the majority religious group tends to perceive a higher threat as the minority grows larger. In such a case, the majority group tries to keep their dominant position by excluding the religious minority from a specific opportunity. Meanwhile, the minority group perceives the competition as unfair (Bobo and Hutchings 1996), which only increases the likelihood of conflicts between the two religious groups. Secondly, perceived threat is also concerned with the fear of losing political control or power (Olzak 2013). This is more likely when there are economic upheavals. As such, the religious majority starts displaying exclusionary measures against the religious minority as an attempt to protect their privileges.

From what we have laid out so far, perceived threat appears to surface at the contextual level and the individual level (Scheepers et al. 2002b). At the contextual level, the threat depends on the macro-social conditions such as the (trans)migration flux in the neighbouring cities, and the meso-social conditions such as ethnic and religious segregation in jobs. At the individual level, the threat depends on the severity of intergroup competition subjectively perceived by individuals. It is important to remember that perceived threat is not limited to economic context but can also be in form of cultural threat. One response often found towards the influx of migrants is the belief that they will eventually change the existing cultural structure (Zárate et al. 2004). This is in line with the claim made by the integrated threat theory. The theory posits that intergroup threat rises when ingroup members perceive their material and symbolic resources being threatened by the presence of the outgroup (Stephan et al. 2000). Our research context also reveals that Christians are not only a threat to the Muslim communities’ realistic resources (e.g., job, land, political votes) but also a symbolic threat to the Muslims’ values as a religion (Hadiz 2017). Based on the foregoing notions, the hypotheses related to perceived threat are formulated as follows:

Hypothesis 7 (H7).

Perceived threat positively mediates the relations between religious practice and beliefs, and supports interreligious conflicts.

Hypothesis 8 (H8).

Perceived threat negatively mediates the relation between religious salience and supporting interreligious conflicts.

3. Materials and Methods

This study used secondary data obtained from a cross-cultural dataset on interreligious conflicts in six conflict regions in Indonesia. The dataset was made publicly and freely available for secondary analyses (Setiawan et al. 2018). Here, we only provide brief explanations of participants and sampling procedures. For further explanation of the dataset, extensive documentation of it can be found in Data Archiving and Networked Services (DANS).

3.1. Participants and Sampling Procedures

The data collection was conducted from May until August 2017 and involved six locations in Indonesia where religious hostilities have been prevalent: Singkil (Aceh Province), South Lampung (Lampung Province), Bekasi (West Java Province), Poso (Central Sulawesi Province), Kupang (East Nusa Tenggara Province), and Sampang Madura (East Java Province). Thus, the dataset covers a vast area of the whole Indonesian archipelago. Figure 1 provides a map of Indonesia to give readers a better understanding of the research locations. The survey aimed to collect a random sample of the general population aged 17–65 years old, with the criteria of having lived in the area for a minimum of five years.

To achieve the aim, we employed a local research team to work together to approach the residents. In doing so, we employed two random sampling procedures. First, whenever we successfully attained the available regional population registry, we conducted a simple random selection by throwing dice to determine the starting point of the household. Prior to random selection, we also calculated the sampling interval by dividing the number of populations in the registry by the number of expected households (i.e., 100 for each neighbourhood). For example, with a sampling interval of 20, when the dice shows number 3 then we start from the household listed in that number and continue to household number 23, then 33, and repeat the calculation until all 100 households are achieved. Second, when the population registry was not available or substantially inaccurate, the survey employed a local random walk. These random selection procedures were employed to reduce biases on the part of researchers and thus, the samples constitute the best approximation of a representation of the full adult populations in those areas (Babbie 1989).

Next, when people inside the household agreed to participate, we explained our selection criteria and randomly selected the adult within the household by the most recently celebrated birth date. Throughout our data collection, we meticulously followed ethical considerations to ensure that respondents were accurately informed about the study and had an option to voluntarily participate in the study. After agreeing with the study terms, we asked participants to fill out the consent form. At the end of the study, participants were given a small gift of approximately €2.

The dataset contains 2356 participants consisting of various religious affiliations. For this study, we only selected Muslim and Christian (both Protestants and Catholic) participants. We also removed participants with a substantial number of missing values, especially the dependent variable. We were able to gather 2026 participants, 1452 Muslims and 574 Christians. Within that number, we have 991 females and 1035 males. For further descriptive statistics, please refer to

Table 1. Descriptive statistics and mean differences.

Variables	Range	Muslims		Christians		t-Test
		Mean	SD	Mean	SD
Lawful protests	1–5	3.40	0.86	2.98	0.94	−9.17
Violent protests	1–5	2.28	0.84	1.86	0.64	−12.20
Religious service attendance	1–7	3.58	1.63	4.02	1.01	7.29
Particularism	1–5	3.99	0.72	3.26	1.05	−15.34
Religiocentrism	1–5	3.18	0.66	2.76	0.75	−11.63
Salience	1–5	4.01	0.85	4.20	0.83	4.51
Memory of conflicts	1–5	1.46	0.78	1.67	0.91	5.01
Perceived threat	1–5	2.72	0.95	2.19	0.74	−13.48
Individual characteristics
Age	17–65	33.07	12.22	31.03	11.49	−3.53
Sex	0/1	0.51	0.50	0.52	0.50	-
Education	1–6	3.47	1.08	3.95	0.97	9.78
Income	1–8	3.50	2.04	4.11	1.96	6.20

Bold indicates significant at p < 0.05 with two-tailed tests.

.

3.2. Measures

We used measures that have been previously studied and tested in a similar research context. However, as previously mentioned, our study involved the general population in the areas of interest. Therefore, we must ensure to run appropriate validity and reliability tests to show the applicability of the measures. For this, we ran confirmatory factor analysis (CFA) to test the validity of the measures using lavaan package in R (Rosseel 2018). As for reliability, we performed a calculation of composite reliability (CR). In addition, we also tested discriminant validity by calculating the square root of variance extracted (AVE) to make sure that each measured construct is different from the other (Fornell and Larcker 1981).

Table 2. Bivariate correlations and AVE.

	1	2	3	4	5	6	7	8
1. Lawful protests	-	0.32	−0.06	0.30	0.24	0.09	0.15	0.22
2. Violent protests			−0.03	0.14	0.28	−0.10	−0.02	0.35
3. Religious service attendance				−0.04	−0.02	0.11	−0.05	−0.03
4. Particularism					0.50	0.16	0.01	0.30
5. Religiocentrism						0.09	0.07	0.42
6. Salience							0.13	0.00
7. Memory of conflicts								−0.01
8. Perceived threat								-
AVE	0.55	0.65	-	0.58	0.34	0.64	-	0.62

Bold indicates significant at p < 0.05.

provides correlation between variables and AVE.

3.3. Supporting Interreligious Conflicts

To measure our dependent variables, we adopted a scale of supporting interreligious conflicts previously used by Subagya (2015). The scale consists of two modes of support, that is supporting lawful protests and supporting violent protests against the religious outgroup. The former mode measures the level of participants’ support for public criticism and demonstrations, while the latter measures the level of participants’ support for damaging of religious outgroup’s property and harming of religious outgroup members. Each scale contains six items and is rated on a five-point Likert scale, with higher scores indicating stronger support for lawful and violent protests (see Appendix A for a full scale).

We ran a multi-group CFA to test a two-factor model of this scale. The results found that χ² = 1126.03, p < 0.00, CFI = 0.93, root mean square of approximation (RMSEA) = 0.10, and the standardized root mean squared (SRMR) = 0.05. Although the RMSEA is slightly above the cut-off (0.07), the rest of the good fit indicators suggest a good fit model (Hu and Bentler 2009; Hooper et al. 2008). Further, the standardized parameter estimates suggest that all items in each mode share a considerable variance with its relevant factor, ranging from 0.69 to 0.79 for supporting lawful protests for Muslims and from 0.72 to 0.81 for Christians. Whereas factor loadings for supporting violent protests range from 0.73 to 0.87 for Muslims and from 0.69 to 0.91 for Christians. Lastly, both scales were found to be highly reliable for both religious groups, CR = 0.87 for Muslims and 0.89 for Christians supporting lawful protests, and CR = 0.91 for both religious groups supporting violent protests.

3.4. Religious Practices

We used a single straightforward question to measure participants’ attendance to religious services, “How often do you go to religious services in mosques, churches, temples or other places of worship?”. The question is rated on a seven-point scale, ranging from ‘never’ to ‘several times a day’.

3.5. Religious Beliefs

Two measures of religious beliefs were used in this study. First, a three-item particularism scale was used to measure the participants’ tendency to view religious doctrines as the absolute truth. The scale was based on the study of interpreting religious plurality by Anthony et al. (2015). We used items such as, “Compared with my religion, other religions contain only partial truths”, and asked participants to rate themselves on a five-point scale (from ‘totally disagree’ to ‘totally agree’). Second, a five-item religiocentrism scale was adopted from a study on religiocentrism scale by Sterkens and Anthony (2008) to measure the extent to which participants show positive attitudes towards the religious ingroup and negative attitudes towards the religious outgroup. Positive attitudes towards the religious ingroup are represented by items, such as “Thanks to our religion, most of us are good people”, and negative attitudes towards religious outgroups are measured by items such as, “Other religions only talk about doing good deeds without practising them”. Similarly, participants were asked to rate themselves on a five-point scale, with higher scores suggesting higher agreement to the items.

Similar to our dependent variables, we ran a multi-group CFA to test the validity of the two-factor model. The good fit indicators, χ² = 407.31, p < 0.00, CFI = 0.91, RMSEA = 0.10, and SRMR = 0.06, suggest that the model fits the data. For particularism scale, each item shows an acceptable level of estimate, ranging from 0.50 to 0.83 for Muslims and from 0.69 to 0.89 for Christians. Likewise, each item in religiocentrism scale shows a good range of factor loading, from 0.49 to 0.66 for Muslims and from 0.53 to 0.75 for Christians. Finally, our reliability analysis suggests that the scales are moderately reliable across religious groups. Specifically, particularism scale has CR = 0.72 for Muslims and 0.83 for Christians. For religiocentrism scale, the CR = 0.67 for Muslims and 0.76 for Christians.

3.6. Salience

Religious salience was measured using a three-item scale from Eisinga et al. (1991). The scale asks participants to what extent their religion plays an important part in their life daily. The scale contains statements such as, “My religious beliefs have a great deal of influence on how I relate with others” and asks participants to rate themselves on a five-point scale.

Our multi-group CFA shows that the model fits the data well, χ² = 30.75, p < 0.00, CFI = 0.99, RMSEA = 0.12, and SRMR = 0.04. Each item shows high level of factor loading, ranging from 0.77 to 0.81 for Muslims and from 0.73 to 0.88 for Christians. In addition, the scale is also shown to be reliable across religious groups (CR = 0.84 for Muslims and CR = 0.86 for Christians).

3.7. Memory of Conflicts

We used a single statement to measure memory of conflicts among participants. The statement used was intended to measure the frequency of discussing interreligious conflicts that have happened in their area. Participants were asked a statement “In your family, how often do you talk about the interreligious violence that happened in your area?” and rated themselves on a four-point scale (from ‘never’ to ‘often’).

3.8. Perceived Threat

This measure was adopted from an intergroup competition study by Scheepers et al. (2002b). It operationalized perceived threat as a subjective perception towards the severity of intergroup competition. In this measure, we focused on the collective interests of the religious ingroup rather than the individual. This is due to the high relevance of religious ingroup narratives regarding collective interests in the making and escalating of conflicts (see van Bruinessen 2018; Hadiz 2017; Human Rights Watch 2013).

We ran a multi-group CFA to demonstrate that the model fits the data, χ² = 23.87, p < 0.00, CFI = 0.99, RMSEA = 0.07, and SRMR = 0.01. The standardized parameter estimates are within a highly acceptable level, ranging from 0.71 to 0.84 for Muslims and from 0.70 to 0.86 for Christians. Finally, the scale is also shown to be reliable across religious groups (CR = 0.86 for both religious groups).

3.9. Individual Characteristics

Straightforward questions were employed to measure age and gender. Next, the data on the level of education indicates their completed highest level of education, ranging from ‘did not go to school’ (1) to ‘Master’s degree or higher’ (6). Finally, we asked for monthly gross household income, ranging from ‘Lower than Rp. 500.000’ (1) to ‘Rp. 6.000.000, and over’ (8).

3.10. Measurement Invariance

To make sure that both Muslim and Christian participants respond in the same way to the items presented, we also conducted analyses of measurement invariance (MI) as part of multi-group CFA (Milfont and Fischer 2010). In detail, we tested the metric invariance to demonstrate that the meaning of items was identical to both groups (Bagozzi and Edwards 1998). To do this, we first set up a configural model to examine whether the same configuration of items exists for both groups. Next, we set up a model which constrained factor loadings to be equal across groups to test the metric invariance. If the chi-square test between the configural and metric model is insignificant, then we can retain our assumption that both groups responded in a similar manner to the latent constructs in our measure (Putnick and Bornstein 2016). However, chi-square is very sensitive to large sample size. Therefore, we also look at the difference in confirmatory fit index (CFI) or ΔCFI between the two models. If the difference is less than 0.01, as pointed out by Cheung and Rensvold (2002), then we can safely say that our measure is invariant across religious groups.

Table 3. Results of measurement invariance testing.

Scale	Differences
	χ2	df	p	ΔCFI	Conclusion
1. Supporting interreligious conflicts	36.90	10	0.000	−0.002	Invariant
2. Religious beliefs	13.12	6	0.041	−0.002	Invariant
3. Religious salience	4.18	1	0.041	−0.001	Invariant
4. Perceived threat	0.47	3	0.925	−0.001	Invariant

provides the full account of measurement invariance testing.

Furthermore, we are also aware of the imbalance of size in our religious groups. The sample size ratio between Muslims and Christians is higher than 2 to 1. However, Yoon and Lai (2018) have demonstrated that the imbalance ratio in our group sizes will not significantly impair our conclusion on MI. Therefore, we can safely conclude that all the measures employed appeared to be invariant across both religious groups.

3.11. Strategy for Analyses

Prior to testing the hypotheses using structural equation modelling (SEM) in lavaan package, we performed three preliminary tests to meet the statistical assumptions. First, we calculated skewness and kurtosis to ensure that our dependent variables follow a normal distribution. Our results show that the values of skewness and kurtosis of the dependent variables are less than 2 and 7 respectively, identifying no substantial departure from normality (Kim 2013). Second, we tested the linearity of the dependent variables and found that they were all linearly distributed. Third, we ran multicollinearity diagnostics to ensure that there will be a highly shared variance among the predictors. For this, we found that the scores of variance inflation factor (VIF) and tolerance statistics of the predictors are within the normal range of less than 10 and more than 0.2 respectively (Field 2009). The non-multicollinearity between predictors can also be seen from Table 2, where there are no high correlations between predictors.

4. Results

We conducted independent sample t-tests to provide preliminary findings on the mean differences between the two religious groups. Table 1 shows that there is a substantial difference between Muslim and Christian participants in their support for interreligious conflicts. On average, Muslims (M = 3.40, SD = 0.86) were found to have higher support for lawful protests compared to Christians (M = 2.98, SD = 0.94), t(973.81) = −9.17, p = 0.00. Similarly, on average, Muslims (M = 2.28, SD = 0.84) were also found to have higher support for violent protests compared to Christians (M = 1.86, SD = 0.64), t(1384.63) = −12.20, p = 0.00. Based on the mean scores of the two modes of protests, we see that both religious groups appeared to be more reluctant to support violent protests. However, the group difference still signaled a strong tendency of exclusionary measures which may be related to a sense of group position held by both religious groups (Olzak 2013).

Next, we move to the relations between variables examined via SEM. Our SEM analysis showed the following fit indices, χ² = 137.21, p < 0.00, CFI = 0.91, RMSEA = 0.08, and SRMR = 0.03. According to Hooper et al. (2008), these results demonstrated a good fit model. Therefore, we can now start exploring the proposed relations. Our first hypothesis stated that religious service attendance is positively related to supporting interreligious conflicts. We found that attendance to religious service is related only to supporting lawful protests, (b = −0.04, p = 0.00), and the relation was negative. Therefore, our hypothesis on religious practices (H1) is fully rejected. Further, we also found that religious attendance is negatively related to the frequency of past interreligious conflicts discussion (memory of conflicts) among participants, (b = −0.04, p = 0.00).

The second and third hypotheses were on the relation between religious beliefs and supporting interreligious conflicts. In detail, we found that particularism is positively related only to supporting lawful protests, (b = 0.22, p = 0.00). By this, we partially accept the claim that particularism is positively related to supporting interreligious conflicts (H2). Subsequently, religiocentrism was found to be positively related to supporting lawful protests (b = 0.09, p = 0.01) as well as to supporting violent protests (b = 0.21, p = 0.01). Here, we see that religiocentrism was much stronger related to the latter support as compared to the former. Based on this, we fully accept the hypothesized relations between religiocentrism and supporting interreligious conflicts (H3). Interestingly, we also found that particularism is negatively related to memory of conflicts, (b = −0.05, p = 0.04), while on the other hand, it was positively related to perceived threat, (b = 0.14, p = 0.00). As for religiocentrism, we found that it is positively related to memory of conflicts (b = 0.09, p = 0.00) and perceived threat towards the religious outgroup (b = 0.47, p = 0.00). Here, not only was the relation consistent with theoretical claims (see Brewer 2001; Wright and Young 2017) but was also much stronger than any relation found so far.

Moving on to religious salience, we hypothesized that there should be a negative relation with both modes of support (H4). However, we found that religious salience is only negatively related to supporting violent protests, (b = −0.09, p = 0.00). Therefore, we partially accept the fourth hypothesis. Further, we found that salience is positively related to memory of conflicts (b = 0.14, p = 0.00), but, as expected, is negatively related to perceived threat (b = −0.05, p = 0.04).

Before moving on to the mediation analyses, it is important for us to also look at the relations between mediators and supporting interreligious conflicts. Our results showed that memory of conflicts is positively related to supporting lawful protests (b = 0.15, p = 0.00), but not to supporting violent protests. Whereas perceived threat was found to be positively related to both modes of protest (b = 0.13, p = 0.00 for lawful protest and b = 0.24, p = 0.00 for violent protest).

From the results so far, we can expect to have several mediated relations appear significant. Figure 2 provides an overview of the hypothesized relations. First, we discovered that memory of conflicts mediates most of the relations between religiosity dimensions and supporting lawful protests. In detail, memory of conflicts, unexpectedly, negatively mediated the relation between religious attendance and supporting lawful protests, b = −0.01, p = 0.00. Further, memory of conflicts positively mediated the relations between religiocentrism and religious salience on one hand and supporting lawful protests on the other (b = 0.01, p = 0.01 and b = 0.02, p = 0.00, respectively). The latter relation was found to be contrary to the one we expected. Earlier we found that memory of conflicts is not related to supporting violent protests. Statistically, the result should make memory of conflicts a non-significant mediator to explain the relations between religiosity dimensions and supporting violent protest. Therefore, we partially accept H5 and fully reject H6.

Next, perceived threat seemed to be a stronger mediator for the relations of interest. Specifically, it was found to positively mediate the relations between particularism and religiocentrism on one hand, and supporting lawful protests (b = 0.02, p = 0.00 and b = 0.06, p = 0.00, respectively). Expectedly, it also explained the relations between the same religiosity dimensions and supporting violent protests, b = 0.03, p = 0.00 for particularism and b = 0.11, p = 0.00 for religiocentrism). Based on this, we partially accept H7 and fully reject H8.

Regarding individual differences, we found that being male, as predicted, is more related to supporting interreligious conflicts (b = 0.18, p = 0.00 for lawful protests and b = 0.10, p = 0.00 for violent protests). Level of education also significantly predicted the support for lawful protests (b = 0.06, p = 0.00), but not violent protests. Conversely, level of income, unexpectedly, significantly predicted a lowered support for violent protests (b = −0.04, p = 0.00).

5. Discussion and Conclusions

In this study, we scrutinized how relations between different religiosity dimensions and supporting interreligious conflicts are explained by relevant mediators, namely memory of conflicts and perceived threat. To do this, we took religiosity dimensions as the manifestation of individuals’ extent of religious identification (Cameron 2004; Stark and Glock 1968). Further, we included two modes of support as latent behaviour which have been reported to positively impact interreligious tension (Human Rights Watch 2013; Halili 2016). As for the mediators, we looked at the frequency of participants discussing past interreligious conflicts in their area as a reference to the memory of conflicts and their perceptions towards the religious outgroup’s threat against their religious ingroup’s collective interest (Bar-tal 2007; Scheepers et al. 2002b). We found a substantial number of significant relations between predictors and outcome variables, corroborating previous studies.

In detail, we hypothesized that (H1) religious service attendance, (H2) particularism, and (H3) religiocentrism are positively related to supporting interreligious conflicts, whereas (H4) religious salience was expected to be negatively related to supporting interreligious conflicts. We found no positive relation between religious service attendance and supporting interreligious conflicts. For particularism, we only found a positive relation with supporting lawful protests. However, we found a positive relation between religiocentrism and supporting interreligious conflicts. Finally, we found that religious salience is partially related to supporting interreligious conflicts, which is negatively related only to supporting violent protests.

Further, we also tested relevant mediators to explain the relations between religiosity dimensions and supporting interreligious conflicts. We hypothesized that (H5) memory of conflicts and (H7) perceived threat positively mediate the relations between religious practice and beliefs and supporting interreligious conflicts. On the other hand, we hypothesized (H6) memory of conflicts and (H7) perceived threat to negatively mediate the relation between religious salience and supporting interreligious conflicts. We found that both mediators are able to significantly predict the support for interreligious conflicts, with perceived threat being a much stronger predictor compared to the memory of conflicts.

Specifically, memory of conflicts positively mediated the relation between religio-centrism and supporting lawful protests. This is in line with previous studies that show ingroup favouritism and negative outgroup attitudes combined are more likely to push individuals to display exclusionary measures against the religious outgroup (Anthony et al. 2015; Brewer 1999). However, as Brewer (1999) pointed out, negative outgroup attitude is not necessarily a by-product of ingroup favouritism. We agree with her because positive attitudes towards ingroup is a requirement for individuals to be able to identify themselves with a religious identity (see Turner 1975). In doing so, individuals are predetermined to value other religious identities less positive, a process where they constantly compare their religious identity to outgroups as a means to maintain their superiority or psychological distinctiveness (Ysseldyk et al. 2010). Furthermore, it is worth noting that there may also be people who were heavily impacted by interreligious conflicts but choose to keep the recollection of the conflicts intact to themselves. Hence, this makes memory of conflict less strong to explain the relation between religiosity dimensions and exclusionary measures against the religious outgroup.

In addition, this is especially relevant when two conflicting religious groups have a prior hostile experience. This memory of conflicts not only connects the religious beliefs regarding ingroup-outgroup, but it can also hinder the emergence of intergroup trust (Bar-tal 2007; Tam et al. 2009). Hence, making conflict resolution more difficult. Although memory of conflicts has the potential for the resurgence of interreligious conflicts, we found more comforting support in which it only pertains to supporting lawful protests. One way to explain this is the participants may have endured great pain during past interreligious conflicts and therefore, are reluctant to talk about supporting physical clashes during their memory sharing of past conflicts among each other. As shown by Braithwaite and Leah (2010) in Maluku and North Maluku cases, it was the local people who took charge in leading peaceful reconciliation without arms, as opposed to military forces, due to their weary and long-ending battles between Muslim and Christian communities. It is also supported by De Juan et al. (2015) who demonstrated that local religious institutions play a big part in pacifying the conflicting groups. Therefore, sharing a memory of conflicts does not necessarily lead to destructive outcomes. This can be a way to share ingroup biases among the communities and thus, leading to better ways of dealing with interreligious competition in both religious groups. However, lawful protest is undoubtedly not to be taken lightly, as it has proven to be capable to transform into violent clashes in the past.

Next, perceived threat positively mediated the relations between particularism and religio-centrism on one hand and both modes of support on the other. Unlike memory of conflicts, perceived threat shows to act as a much stronger mediator in the relations. This conforms to two important points. One, a religious outgroup is considered to hold differing worldviews and this belief is amplified by the fact that most religious adherents are expected to accept their religious truth as the only supernatural truth (Stephan et al. 2000; Stark and Glock 1968). Thus, living side by side with a religious outgroup imposes a constant threat to their religious norms and beliefs, especially when the outgroup is growing eminently. Two, based on proposition one, own religious beliefs provide individuals with certain stereotypes about the ingroup as well as the outgroup. These stereotypes presuppose them with prejudice against the religious outgroup (Duckitt 2003). Combined with feeling threatened by the religious outgroup, this prejudice and other exclusionary measures then become the main defence to perceived threat (Olzak 2013; Quillian 1995).

In sum, perceived threat not only connects religious identity (manifested by religiosity dimensions) with supporting interreligious conflicts, but it is also an inevitable response to living side by side with a religious outgroup. However, this should also be taken proportionately. According to Coser (1956), intergroup competition (and even conflicts) is normal for groups sharing the same living habitat. In fact, intergroup competition is needed to keep ingroup members cohesive and cooperative. Blalock (1967) added that intergroup competition is also useful to keep group members aware and maintain or improve their ‘superiority’. Putting it in a broader sociological context, interreligious competition is somewhat needed to keep all society members improving themselves and innovating to achieve a greater good. This way members from both religious groups have a higher chance to interact and diminish prejudice held against each other (Savelkoul et al. 2014; Pettigrew et al. 2007). Therefore, perceived threat is a constant reminder that future interreligious conflicts still may erupt. To prevent this, members from both religious groups should work together in their competition to achieve a ‘highly functional’ conflict.

Regarding the overall findings, we conclude that the study offers a unique perspective in looking at interreligious conflicts in Indonesia. First, the positive relation between religio-centrism and supporting interreligious conflicts confirms the claim that religiosity is still considered one of the most important traits among Indonesians in a new democratic period of the country (Hadiz 2017). Although the claim sounds benign to a lot of people, it may be harmful to the progress of democracy in the country. A high degree of religiosity, especially religio-centrism, encourages people to involve religious ingroup favouritism in every aspect of their life. In addition, with the growing perceived threat towards the religious outgroup, religio-centrism can easily be associated with exclusionary measures against the religious outgroup. We can use this relation to explain why there have been growing demands to impose specific religious regulations in handling socio-political matters in various cities in Indonesia. For instance, local Sharia regulations in several areas and the Jakarta gubernatorial election in 2017. Meanwhile, the study also confirms a more universal perspective in evaluating interreligious conflicts. By being the strongest mediator in this study, perceived threat confirms itself as the main driving factor to collective action (Blalock 1967; Quillian 1995; Olzak 2013).

We acknowledge that our study can be improved in many ways. First, due to the cross-sectional nature of the dataset, future studies are encouraged to look at the different periods in relation to participants’ perceived threat and supporting interreligious conflicts. In Indonesia, and applies to most of the world, interreligious competition is often used as a narrative to gather political votes (Hadiz 2017; van Bruinessen 2018). Therefore, it is wise to compare data from non-political times to further disentangle the impact political competition brings on supporting interreligious conflicts. Second, as previously discussed, local context plays a big part in escalating as well as de-escalating conflicts (De Juan et al. 2015; Barron et al. 2009). Therefore, involving regional macro-variables, e.g., minimum income of province, regional religious composition, has high potential to further explain the complexity of interreligious conflicts. Third, as previously mentioned, our group sample sizes were not equal. While it is true that most provinces in Indonesia are predominantly Muslims, future research can still improve their research by including more Christian-populated areas to obtain a better ratio of Muslims and Christians. This way they can improve the statistical power when comparing two religious groups. Fourth, we acknowledge the limitation of the memory of conflicts measure. By using this measure, we could not capture people who keep silent about past conflicts but may hold strong memory of the conflicts. Therefore, we encourage future research to consider a more in-depth qualitative study to explore the contents of memory of conflicts among those who have lived through such experiences. Finally, scholars are encouraged to pursue an experimental route to explain the potential causal relations mediated by a memory of conflicts and perceived threat. News regarding religious competition can be used to prime perceived threat and participants are later measured in terms of their support for interreligious conflicts. Experimental research may offer a new perspective to the current literature, as the subject is mostly studied in field research.