Indonesian Civilians’ Attributions for Anti-Chinese Violence During the May 1998 Riots in Indonesia

The Ultimate Attribution Error in Intergroup Relations

Attribution “errors” (or "bias", see Kruglanski & Ajzen, 1983) involve differences in perceptions regarding the causes of events based on one’s social position. For example, the “fundamental attribution error” is the tendency to underestimate the role of situational factors and overestimate personal factors as causes of other people’s behaviour (Heider, 1944; Ross, 1977), compared to explanations for one’s own actions. For example, when a colleague is fired from work, I may blame that individual’s personal failings (attribution to the internal characteristics of that person). In contrast, if I am fired, I believe myself to have been the victim of unfair treatment (attribution to external factors).

Pettigrew (1979) extended this attribution error or bias to the group level, calling it the “ultimate attribution error”. In the ultimate attribution error, often individuals attribute a socially undesirable act of an ingroup member to external factors, while an undesirable act of an outgroup member is attributed to internal factors (see Hewstone, 1990; Pettigrew, 1979). This group bias is consistent with social identity theory, in which individuals experience a shared collective identity that shapes their perceptions and actions, based on their membership in a group (Tajfel & Turner, 1979).

Previous research on collective violence in Western contexts suggests that there are different identity concerns and psychological needs for victim and perpetrator groups (Bilewicz & Vollhardt, 2012; Hirschberger, 2018; Vestergren, Drury, & Hammar Chiriac, 2019). Perpetrator group members are more concerned with protecting their in-group’s moral identity and avoiding responsibility and consequences, while victim group members are concerned with protecting their in-group from further harm and having their victimisation acknowledged (Bilali & Vollhardt, 2019). A range of studies have demonstrated this ingroup and outgroup bias in other contexts, showing that the UAE occurs in the judgements of groups with a history of intense conflict and negative stereotypes, or in those communities which experience high levels of prejudice, and are perceived to be socioeconomically advantaged (Doosje & Branscombe, 2003; Hewstone, 1990; Hunter, Stringer, & Watson, 1991).

Our study contributes a valuable new case (Pribumi/ethnic Chinese violence in Indonesia in 1998) to the very limited research conducted on UAE in non-WEIRD contexts. The few studies conducted about UAE in non-WEIRD contexts, such as Khan and Liu (2008); Khandelwal, Dhillon, Akalamkam, and Papneja (2014); Misra and Mishra (2020), did not fully support the UAE. In-group members did not necessarily attribute out-group members’ harmful actions to internal factors, and thus no intergroup bias was found in their research. Thus, further research into intergroup attribution bias within other non-WEIRD contexts is needed, particularly in contexts where there is a history of conflict. In our study, where there is history of discrimination and conflict between Pribumi and Chinese Indonesians (Melvin, 2013; Tan, 2008), we hypothesized that if a participant were Chinese Indonesian (belonging to the victim group), s/he may attribute the causes of the mass violence during the riots to the internal factors of perpetrators. On the other hand, if a participant were Pribumi Indonesian (belonging to the group involved in perpetrating the violence) s/he may attribute the causes of the violence to external factors.

The UAE is interpreted as motivated bias, whereby groups in the same conflict have different psychological representations of the same event (see Bilali & Vollhardt, 2019). However, another explanation for bias could be differing sources of information. For example, people in the victim group may have formed their understandings about these events based on personal experiences or word of mouth from family and friends, whereas people from the perpetrator group may have obtained their information from media stories. The present research explicitly tests this for the first time, examining whether the source of information was a mechanism that reinforces group differences in attribution bias for the May 1998 riots.

In addition to examining perceptions of the causes of the violence, the present research examines perceptions of the extent of the violence, in terms of the number of victims. At the interpersonal level, victims perceive more harm compared to the participants belonging to the perpetrator group (see Baumeister & Hastings, 2013; Bilali, Tropp, & Dasgupta, 2012). Similarly, at the intergroup level, the in-group’s victimisation may be categorized as more severe than other groups’, sometimes termed “competitive victimhood” (Noor, Vollhardt, Mari, & Nadler, 2017). In the present research, we explored whether group differences in evaluations of the riots would be observed in victim estimates, so that Pribumi would minimise the number of victims compared to Chinese Indonesia participants.

Research Context

The riots of 12–15 May 1998 occurred against a backdrop of great political and economic unrest in Indonesia, with a student-led protest movement calling for an end to the military regime (Aspinall, Feith, & Van Klinken, 1999). At the height of these demonstrations in early May 1998, military leaders sought to deflect anger by scapegoating Indonesia’s ethnic Chinese minority, blaming them for the economic crisis crippling the country (Purdey, 2006; Tim Relawan Untuk Kemanusiaan, 1998/1999).

The Indonesian government undertook a partial investigation months after the riots which did little to shed light on the violence or the actors involved (Tim Gabungan Pencari Fakta, 1998). An in-depth civil society-led investigation undertaken shortly after the riots (see Tim Gabungan Pencari Fakta, 1998), in comparison, found that the violence had been incited and led by military actors but with widespread civilian involvement, and estimated that at least 1,000 people had been killed, and an unknown number of Chinese women and girls raped (Tan, 2008; Wandita, 1998). To explain the violence, some researchers have drawn attention to the simultaneous violent attacks across multiple areas of the capital that would have required military organisation and training (Purdey, 2006; Siegel, 1998; Strassler, 2004). Others have also highlighted a longer history of discrimination and violence against the Chinese community to contextualise the riots (Chua, 2007; Purdey, 2006).

Based on the existing research, we identified a range of possible attributions for the causes of the riots, which we classified as either internal or external. The Asian financial crisis, anger at the government, and anger at the police/security services were categorised as relevant external or situational causes, because they are attributable to the political crisis and how the military regime responded. Anti-Chinese prejudice by the Pribumi community, criminal motives (e.g., desire to loot and rape), and government scapegoating of Chinese Indonesians (if the government is seen as representative of Pribumi rather than an oppressive outside force) were categorised as internal or dispositional causes, because these are attributable to the characteristics and actions of the perpetrators themselves. We are also open to new themes that could describe the attributions for the riots, so a mixed-method approach was used in this study. Quantitative measures were used to examine previously identified themes, while qualitative sections of the survey allowed participants to provide additional information, pointing to new themes and enriching insights into their perceptions.

Intersectionality of Ethnicity and Gender

While exploring perceptions of the riots in relation to ethnicity is clearly relevant, as a second contribution of the paper, we draw on the proposition that in intergroup attributions, it is also useful to look at how intersectionality plays a role in influencing how people think and relate to their ingroup and outgroup members (Crenshaw, 1991; Warner et al., 2018). Intersectionality as a framework draws attention to the extent to which no single aspect of identity (e.g., ethnicity) should be studied in isolation without reference to other aspects of identity (e.g., gender) (Collins, 1990).

In this paper, we look at the intersectionality of two highly salient, interrelated, social identities that were critical both during the riots, and to the perceptions of what caused the violence: ethnicity and gender. Both identities are important: ethnicity because of the longstanding ethnic prejudice against Chinese Indonesians and the violence directed towards them as a group during the riots; and gender, because a major element of the violence was the deliberate targeting of Chinese Indonesian women and girls for sexual assaults.

In the quantitative section, we hypothesize that the participants’ ethnicity and gender will affect responses primarily in relation to memberships in the respective perpetrator and victim groups. Overall, we expect to find an effect of participant’s ethnicity, such that Pribumi participants are more likely to minimize the events, and to attribute them to situational factors than Chinese Indonesian participants. We expect this effect to be magnified for Pribumi men, in a gender x ethnicity interaction. Similarly, we expect to find that Chinese Indonesian participants are more likely to attribute the events to internal or dispositional factors of perpetrators, and we expect that this effect will be magnified for Chinese women. While on the qualitative section, we expect to see how the narratives of gender and ethnicity appeared when participants talk about general and sexual violence during the riots.

Participants

Participants in this study were Indonesian adults who met the following criteria: 1) Indonesians who were in Indonesia at the time of the riots; and, 2) over age 30, since this event happened 21 years before the survey was administered. This survey included qualitative and quantitative components. Only participants who completed at least 90% of the qualitative and quantitative components were included in the analysis. For the quantitative components, there was a total of 228 participants (M_age = 41.11, SD = 9.13). From the sample, 56.14% (n = 128) were Pribumi and 43.86% (n = 100) were Chinese Indonesians; 46.05% (n = 105) were male, 53.51% (n = 122) were female, and 0.44% (n = 1) preferred not to say. For the qualitative components, there were 235 participants (M_age = 41.07, SD = 9.07), of whom 57.02% (n = 134) were Pribumi and 42.98% (n = 101) were Chinese Indonesians; 45.96% (n = 108) were male, 53.62% (n = 126) were female, and 0.43% (n = 1) preferred not to say. We recruited participants until we had reached a minimum of 100 Pribumi and 100 Chinese Indonesians, basing our sample size on previous research (e.g. Khan & Liu, 2008).

Design

Quantitative analyses examined perceptions according to participant and target ethnicity and gender, and qualitative analyses were based on Braun and Clarke’s (2019) thematic analysis. The qualitative data complements the information in the quantitative data, and considers how people explain the causes of the riots in more detail within the particular socio-economic and political context of the riots. This mixed-method study aims to give understanding to a complex phenomenon with statistical analyses as well as to explain the phenomena qualitatively (Creswell, 1999).

Procedure

Ethical clearance was obtained from the ethics committee at The University of Queensland. An online survey was conducted with snowball sampling over a period of two months (March–April 2020). The recruitment material and survey were drafted in English, then the lead author translated them into Indonesian, and the third author reviewed the translation. A link to the online survey was shared through emails and various social media platforms (Facebook, WhatsApp, Instagram), asking for voluntary participation through the authors’ personal contacts.

When a person wished to participate, s/he clicked on the link, which then opened the online survey as a separate URL. The first page of the survey contained the participant information sheet and consent form, which explained the background and study purposes, risks associated with the project, voluntary participation, the possibility for withdrawal, and data confidentiality. Participants first completed the demographics questions, then provided qualitative data, and finally completed the quantitative questions. All data were collected and recorded anonymously. Most participants took between 15 and 30 minutes to complete the survey, depending on the level of detail they chose to include in the open-ended responses.

Measures

Quantitative Findings

The Estimated Number of Chinese and Pribumi Victims by Participant Ethnicity and Gender

A 2 (participant gender: male/female) x 2 (participant ethnicity: Pribumi/Chinese Indonesian) x 2 (victim ethnicity: Chinese Indonesian/Pribumi) x 2 (victimisation type: killed/raped) ANOVA was conducted, and revealed a significant main effect of the type of victimisation, F(1, 218) = 206.661, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .487, as well as a significant main effect of victim ethnicity, F(1, 218) = 335.294, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .606, and a significant main effect of participant ethnicity, F(1, 218) = 4.697, p = .031, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .021. As can be seen in Table 4, the main effect of victimisation type reveals that there were more victims perceived to have been killed than raped. In addition, the main effect of victim ethnicity reveals that there were more Chinese Indonesian victims perceived than Pribumi victims, and the main effect of participant ethnicity reveals that the Pribumi participants perceived more victims overall than the Chinese Indonesians did. These main effects, however, were qualified by significant interactions between victim ethnicity and participant ethnicity, F(1, 218) = 22.913, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .095, as well as victimisation type and victim ethnicity, F(1, 218) = 5.042, p = .026, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .023.

Table 4

Means and Standard Deviations for the Estimated Number of Victims by Participant Ethnicity and Participant Gender, as well as Victim Ethnicity and Type of Victimisation

Participant Ethnicity / Gender	Chinese killed		Chinese raped		Pribumi killed		Pribumi raped
	M	SD	M	SD	M	SD	M	SD
Pribumi
Male	3.95	1.10	3.36	1.05	3.20	1.35	2.41	1.16
Female	3.73	1.16	3.14	1.17	3.02	1.09	2.09	1.02
Chinese
Male	3.98	0.98	3.38	0.85	2.57	1.19	1.79	0.81
Female	3.89	0.99	3.18	0.77	2.40	1.07	1.75	0.84
Total	3.88a	1.07	3.26b	0.99	2.83c	1.21	2.04d	1.02

Note. Response range = 1–5 (1 = None; 2 = < 100; 3 = 101–500; 4 = 501–1000; 5 = > 1000). Means with different superscripts within a row indicate significant differences (p < .05).

The gender main effect was not significant, and there were no other significant interactions, Fs < 1.748, ps > .187, ${\mathrm{\eta }}_{\mathrm{p}}^2$ < .009. The simple effects of victim ethnicity are discussed below and the interactions are plotted in Figures 1 and 2.

Click to enlarge

Figure 1

Mean Rated Estimate Number of Victims as a Function of Victim Ethnicity and Participant Ethnicity

Note. Error bars represent standard errors. Response range = 1–5 (1 = None; 2 = < 100; 3 = 101–500; 4 = 501–1000; 5 = > 1000).

Click to enlarge

Figure 2

Mean Rated Estimate Number of Victims as a Function of Victim Ethnicity and Type of Victimization

Note. Error bars represent standard errors. Response range = 1–5 (1 = None; 2 = < 100; 3 = 101–500; 4 = 501–1000; 5 = > 1000).

Decomposing the interaction of participant and victim ethnicity, simple effects of victim ethnicity were significant for both Pribumi respondents, F(1, 218) = 105.695, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .327, and Chinese Indonesian respondents, F(1, 218) = 235.077, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .519. In both cases, more Chinese Indonesian victims were perceived than Pribumi victims, although the effect size was larger for Chinese Indonesian participants than Pribumi participants.⁴ Similarly, in decomposing the interaction of type of victimisation and victim ethnicity, the simple effects of victim ethnicity were significant for both estimates of murders, F(1, 218) = 187.305, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .462, and estimates of rapes, F(1, 218) = 345.956, p < .001, ${\mathrm{\eta }}_{\mathrm{p}}^2$ = .613. Again, more Chinese Indonesian victims were perceived than Pribumi victims, although the difference was larger for the estimates of rapes than the estimates of murders.

Qualitative Findings

Our approach is a realist perspective associated with thematic, narrative-based analysis, and we generated the codes for our qualitative analysis both deductively and inductively (Braun & Clarke, 2019). We ensure the quality of our analysis through verification and reviewed by the third author who is fluent in Indonesian. Data analysis adhered to the process described by King, Horrocks, and Brooks (2018), which includes: descriptive coding, interpretive coding, and the identification of overarching themes. The survey was conducted in Indonesian language and the first coding was performed by the lead author. The first descriptive coding involved grouping responses based on semantic similarities. The next phase involved the analysis of similar patterns or responses to develop categorical codes. This phase was reviewed by the third author. Based on the categorical codes, key themes were then identified. All members of the research team reviewed the codes and themes to ensure comprehensive coverage and accuracy. Differences were resolved by discussion.

The qualitative findings on how participants attributed the causes for the May 1998 riots affirmed the quantitative findings. As in the quantitative data, anger toward the government was most frequently reported as a cause (external), followed by anti-Chinese prejudice (internal). Other external causes identified were anger with the security services, as well as anger over how the government had handled the financial crisis. Another internal cause (attributable to the characteristics and actions of the perpetrators themselves) was the criminal motives of groups of people who acted violently.

Two novel themes that go beyond and deeper than the quantitative findings were identified in the qualitative responses: military provocation (external) and the vulnerability of some civilian perpetrators to this provocation (internal). Some participants expressed their belief that the military had incited and mobilized civilian participants, pointing to the organisation or coordination of violent actions during the riots. As one Pribumi man explained, “only those with a military background could do this” (50, male, Pribumi). While some participants explicitly named military actors or soldiers, other used the type of euphemistic terms for military actors commonly used in Indonesia to obfuscate agency, particularly when discussing politically dangerous topics (Anderson, 2001). One participant, for example, explained, “On 13 May 1998 at around 10am, they started burning vehicles and tires on the roads, and the shops owned by Chinese people were looted. There was looting and arson, you could see these men with short haircuts committing acts of violence” (53, female, Chinese). Here, “men with short haircuts” is code for “soldiers”.

The vulnerability of civilian perpetrators to incitement was identified as an internal cause, as some respondents stated that this vulnerability was due to deprivation and ignorance, or else was the result of civilians being easily provoked because of their poor economic conditions or lack of knowledge. The idea that community disadvantage is a source of rioting is found in the research on rioting (see Lewis et al., 2011). For example, one woman explained that, “The community was easily triggered to become rioters, the community was also not critical of incitement from other people” (34, female, Pribumi). A Chinese man also commented on these causes, saying “There were parties who deliberately played upon the confusion of the situation and also the educational factors of the Indonesian people which are very low (lack of experience, easily fooled)” (38, male, Chinese).

Ethnicity and Gender

In this qualitative data, the respondents focused on narratives of ethnicity. No gender narratives appeared here, and none of the women or men mentioned themes of male violence/domination. Even when people talked about rape, they focused on ethnicity or criminality. For example, one participant commented, “My assumptions are based on what I remember of that time and what I read, the riots happened because they were instigated by provocateurs who provoked the masses, resulting in looting and rapes, especially against particular ethnic groups (the Chinese)” (32, male, Pribumi). Another stated that “Many ethnic Chinese women experienced sexual violence and were killed at that time. Even one of the witnesses to the mass rapes was killed brutally. The hatred against the Chinese was strongly felt during the May event” (52, female, Pribumi).

Strengths, Limitation and Future Directions

One limitation of this study relates to the use of self-report data from convenience sampling, which may introduce biases in responding and limit the generalisability of the data to the general population. However, to the best of our knowledge, the present study is also the first documentation of Indonesian civilian perceptions of the May 1998 rioting. Our research identified differences in the perceptions of Pribumi and Chinese Indonesians, establishing a novel application of intergroup bias in an under-studied (non-WEIRD) population. Due to the sensitive and taboo nature of the topic, future research could try to triangulate self-report accounts with different measures, such as counts of destroyed buildings in Chinese and Pribumi neighbourhoods in satellite images from more recent rioting, and content analysis of news or social media posts.

To elaborate on this point, while anonymity was maintained in the present research through the use of online surveys, which is beneficial to study such sensitive topics, the reality is that it is still considered dangerous in Indonesia to report some narratives of violence, especially violence perpetrated by state actors such as the military and police. It would not be surprising if participants tend to censor their expression of belief in these narratives. To overcome this self-censorship, future research could use more subtle and indirect probes, perhaps using verbal face-to-face interviews, rather than online data collection.

It is a strength of the present research to examine the intersectionality of gender and ethnicity, but future research may also consider other ways in which intersectionality in the Indonesian context manifests. For example, socioeconomic classes and religious affiliations may play a role in shaping interpretation of the May 1998 riots, or indeed of other events. We could also explore when and how a feminist lens would come into play, and for which types of participants and events.

The impact of age differences is also an interesting direction for future research. The youngest participant in our study was 8 years old at the time of the riots. Events experienced over the age of 6 can produce relatively stable memories (Thompson et al., 2013), especially recollections of extraordinary events such as crimes or violence, which can be more vivid and long lasting (McGaugh, 2003). However, given that 28% of our participants experienced the riots before reaching the age of puberty (mean age of puberty: 11 for female, 12 for male; Blakemore, Burnett, & Dahl, 2010), this may have reduced the impact of gender on them as a lens for the experience. The analysis of age differences would therefore be beneficial for future research, as it could be that women and men who experienced the riots over the age of puberty would be more sensitive to gender-based violence, so that gender identity would play a stronger role in shaping their perceptions.

It is a limitation of the study that we did not validate whether all participants in the victim and perpetrator group interpret the causes as internal and external factors, in line with our approach. For example, a cause such as “government scapegoating of Chinese Indonesians” might be interpreted as external to the perpetrator group by some respondents, but internal by others, depending on whether the government is seen as representative of Pribumi or an oppressive outside force. Future research could include direct questions about causes internal to the Pribumi community and external to it, for a more precise measure of the UAE.

Lastly, regarding the term “Pribumi”, our use of this broad category is not intended to obscure the rich diversity of ethnic and cultural groupings in Indonesia. We acknowledge that future research could profitably engage the distinctions between different ethnic groups which might each have their own distinct historical-political connections to the Chinese Indonesian and Indigenous peoples, such as Arab or Indian Indonesian communities.

Conclusion

The present research examines Indonesian civilians’ perceptions regarding the extent and causes of the anti-Chinese violence in the May 1998 riots. It also explores the intersectionality of ethnicity and gender as factors associated with perceptions. This study contributes to the literature concerning the UAE, and to the psychology of intergroup relations in non-WEIRD contexts in particular. In Indonesia, there have been few studies on the May 1998 riots from psychological perspectives, such as focusing on the traumatic experiences of ethnic Chinese women who experienced rape (Ahmadi, 2021), women survivor’s coping strategies and reactions toward sexual violence (Sutrisno, 2002), and Chinese ethnic stereotyping and discrimination (Kosasih, 2010). This study further makes an important contribution to understanding the attribution bias between groups based on their social identity, in this case ethnicity.

Previous non-WEIRD studies did not fully support the UAE (Khan & Liu, 2008; Khandelwal, Dhillon, Akalamkam, & Papneja, 2014; Misra & Mishra, 2020), while this present research supported the attribution bias between victim and perpetrator groups. This could be due to the long history of anti-Chinese violence in Indonesia, where the New Order government’s policies promoted intergroup prejudice and conflict (see Chua, 2007; Purdey, 2006). In this case, the socio-political context played a critical role in creating attribution bias between groups.

This study also highlights that the UAE appears on a dimension that is salient for the conflict (ethnicity) but not for all groups involved in the conflict (e.g., gender). The influence of media on how respondents frame the events may have an important contribution. Narratives of May 1998 seem to have been predominantly constructed around the ethnicity, rather than the gender, of the rioters and victims.

This study of attributions for mass violence in the May 1998 riots is part of a growing body of intergroup relations research in Indonesia. It enhances the understanding of important psychological processes in intergroup relations that affect the different understandings of perpetrator and victim group members. Such different perceptions may play a role to block reconciliation, by constructing diverging blame narratives and reinforcing the lack of accountability of the perpetrator groups. As different understandings of the events could impede the reconciliation between groups in conflict, addressing the impact of intergroup differences is important for both scholars and practitioners.