Perceived threat from minorities decreases with increased societal threat after the 2021 military coup in Myanmar

Abstract

Societies are often comprised of majority group members who feel threatened by minorities, which, in return, are denied equal rights. How do perceived societal threats, which impact both majorities and minorities, influence perceived minority threat and the support for their rights? We utilized the February 2021 Myanmar coup (Burma)—which has been perceived as threatening by most majority and minority citizens—to examine this question in a three-sample cross-sectional survey. A first sample was conducted before the coup. The second (immediately after) and third (one year after) were aimed to understand short- and long-term differences in perceived minority threat following the coup. Perceived minority threat decreased after the coup with increased perceived societal threat (sample 2) and increased back again when perceived societal threat diminished (sample 3). Lowered perceived minority threat was also associated with higher support for minority rights. Our results reveal the dynamic nature of group-relevant threats using a unique non-WEIRD sample.

Introduction

On February 1st, 2021, a military coup occurred in Myanmar (Burma). The name is widely contested, as Burma is more associated with the dominant ethnic group and past British rule, while Myanmar came forth from the military rule¹. We decided to use “Myanmar” as it better reflects the multi-ethnic structure of the country.)). Myanmar’s military, known as the Tatmadaw, seized control of the country and detained several top political leaders. The coup marked a major setback for Myanmar’s fragile transition to democracy, which had begun in 2011 after decades of military rule². The military’s actions were met with widespread condemnation, and protests erupted across Myanmar, with citizens demanding the release of detained leaders, the restoration of civilian rule, and a return to democratic governance. The military junta responded with a crackdown on dissent, deploying security forces and imposing internet restrictions².

In parallel, Myanmar houses multiple religious and ethnic groups in delicate balance, with Buddhists (i.e., the majority) perceiving ethnic and religious minorities, and especially Muslims, as threats to the identity of Myanmar^3,4.However, the onset of the coup bore far-reaching implications for the freedoms of the whole population, irrespective of ethnicity or religion, as mass population displacement and extreme violence unfolded. Therefore, most Myanmarese citizens, except for those directly affiliated with the military, have seen the coup as a threat to the entire society⁵. This lead majority and minority groups in Myanmar to align for a common cause, sparking joint protest among former rival ethnic groups for the first time since Myanmar’s independence⁶.

Given the tense relationship between the Buddhist majority and the minorities in Myanmar, the coup presents a unique opportunity to examine a long-standing question: what is the relationship between perceived threats to the entire society (e.g., a coup) and perceived threat from minorities experienced by the majority group? The literature suggests two competing hypotheses. The first hypothesis argues that perceived societal threat is associated with increased perceived minority threat and decreased support for minority rights. The second hypothesis argues that a shared perceived societal threat may be associated with lower perceived minority threat and higher support for minority rights. Our study aimed to examine these two competing hypotheses using a large-scale three time-point cross-sectional design in a real-world context.

Perceived societal threat positively associated with perceived minority threat

Numerous historical threatening events involved minorities being scapegoated, i.e., being wrongfully, or disproportionately assigned blame for the events⁷. For example, during the middle-ages, conspiracy theories considered Jews responsible for the spread of the black plague in Europe⁸. More recently, minority groups were blamed and discriminated against during the 2008 economic crisis^{9, 10–11} and the COVID19 pandemic^{12, 13–14}.

Conditions of threat (e.g., during societal threats) may increase the salience of ingroup identities in an attempt to regain control over one’s situation and avoid exposure to additional risks^{15, 16–17}. More particularly, societal threats may shake up the social ranking of societies, threatening the status of advantaged members and making minority members of lower status more threatening¹⁸. In response, individuals may use scapegoating (i.e., assigning blame to minorities for the societal threat) to further enhance one’s control in the face of uncertainty and assuage a sense of individual responsibility⁷. Therefore, minorities- even when they are unrelated to the societal threat- may be seen as more threatening when societal threats arise^{19, 20–21}, further leading to disregard of their civil rights^15,17,22. Based on these findings, one potential hypothesis is that perceived societal threat would be positively associated with perceived minority threat, further relating to lower support for minority rights.

Perceived societal threat negatively associated with perceived minority threat

In contrast to the previously reviewed findings, alternative historical and empirical evidence suggests that threatening situations increase affiliation²³, implying a potential negative relationship between perceived societal threat and perceived minority threat. Research analyzing past disasters and mass emergencies have shown that those events often create shared goals²⁴, therefore being associated with social cohesion, cooperation, and prosocial orientation across groups in society^{25, 26–27}. Furthermore, international threat from China and Russia towards the U.S. has been associated with lowered animosity between American Democrats and Republicans²⁸. One common limitation in these studies, however, is that they assessed the outcomes in the presence of an actual threat (i.e., not necessarily perceived). Therefore, it is impossible to conclude whether perceived threat or another element associated with this threat (e.g., higher contact due to cooperation) underlies the effect on intergroup animosity. In addition, these studies did not focus on perceived threat from minorities.

The common ingroup identity model²⁹ can underpin the social psychology behind the positive effects of perceived societal threat on the relationship between majority and minority groups. The theory posits that people’s identification with various groups (e.g., their homogenous ingroup vs. the broader ingroup including minorities) shifts fluidly depending on whether the context makes salient individual versus collective goals³⁰. According to the model, perceived shared threats make the broader ingroup identity more salient, increasing shared identification with and decreasing threat from individuals who share a superordinate identity^{31, 32–33}. As a result, the second potential hypothesis suggests that as perceived societal threat increases, perceived minority threat decreases, ultimately relating to higher support for minority rights.

The present study

Utilizing the occurrence of the military coup, we designed a three-sample cross-sectional survey in Myanmar to test whether perceived societal threat is positively or negatively associated with perceived minority threat. The first sample was conducted before the onset of the coup, while the second and third samples were conducted one month and one year respectively after the coup. We examined perceived threat from minorities at each sample and measured perceived societal threat and support for minority rights at sample 2 and 3. While sample 1 and sample 2 provided us pre- and post-coup evidence on perceived minority threat, sample 3 allowed us to study the (cross-sectional) relationships between perceived societal and minority threat at a time when the immediate perceived threat from the coup wanes.

The coup followed an election that took place after sample 1 (November 8th, 2020), catching citizens by surprise⁵. Therefore, we assumed that perceived societal threat would be higher after the coup (sample 2 and sample 3) than before the coup (sample 1). Ever since the coup occurred, nationwide long-lasting violent conflict has continued to cause deaths and people to flee their homes³⁴. However, while the number of violent police interventions increased between sample 2 and sample 3³⁵, it was paired with a decreasing- not increasing- public interest as measured by web searches³⁵. In a similar vein, research shows that individuals who live in societies immersed in a chronic state of threat go through routinization, yielding a sense of normalcy in the presence of hardships, risks and other negative consequences³⁶. Given that most of Myanmar’s history is characterized by military rule², we reasoned that in the year after the coup, citizens would habituate to the coup, implying lower perceived societal threat at sample 3 versus sample 2.

The structure of our data allowed us to compare the two competing hypotheses mentioned above. While along with the military coup, a range of psychological variables may alter perceived minority threat, we focus on the perceived societal threat the coup may cause. According to the first hypothesis, perceived minority threat should be higher at sample 2 (post-coup, when perceived societal threat is high) relative to sample 1 (pre-coup, when perceived societal threat is low) and lower in sample 3 (after habituation) relative to sample 2 (H1a). According to the second hypothesis, perceived minority threat should be lower at sample 2 relative to sample 1 and higher at sample 3 relative to sample 2 (H1b). Furthermore, the latter two samples allowed testing the effects of reported perceived threat on support for minority rights. As a result, the first hypothesis argues that the negative relationship between reported perceived societal threat and support for minority rights will be mediated by higher perceived threat from minorities (H2a), while the second hypothesis suggests that the positive relationship between reported perceived societal threat and support for minority rights will be mediated by lower perceived threat from minorities (H2b).

Results

Hypothesis 1

We first tested H1a and H1b, to examine whether perceived minority threat would be lower or higher directly after the coup versus before the coup (occuring right before sample 2 was collected), as well as whether perceived threat would differ a year after the coup relative to right after the coup. We statistically compared the differences in perceived ethnic minority threat and perceived Muslim threat before (Sample 1, when the threat of coup was low), right after the coup (Sample 2, when the threat of coup was high) and a year after the coup (Sample 3, when we assumed salience of the coup would be lower) in two separate models.

Each model consisted of an ANCOVA in which perceived minority threat was predicted by sample and the demographic variables. Gender (0 = “man”, 1 = “woman”), city-resident (0 = “Living in a town or a village”, 1 = “living in a city”) and religion (0 = belonging to a minority religion, 1 = Buddhist, majority religion) were dummy-coded before being entered into the model. The assumptions of the models reported in this study were assessed and deemed satisfied after visual inspection of the assumption plots. For the intersample comparison of perceived minority threat and perceived Muslim threat, the predictor variable “sample” had no missing values, but missingness of the control variables always coincided with dependent variable missingness. This suggests that the missingness is not at random (NMAR, Rubin³⁷). Therefore, both PMM and MCMC methods were used to model intersample differences (Rubin³⁷). Specifically, we employed the mice package in R (Buuren & Groothuis-Oudshoorn³⁸; see Supplementary Materials p. 2–3 for details on the imputation process).

For perceived ethnic minority threat (see Fig. 1 below), two type III one-way ANCOVAs (one per imputation method) examined the effect of sample on the dependent variable, controlling for gender, age, education, religious affiliation (Buddhist vs. other religions), and city residence (see Table 3 and 5, SM). There was a small significant effect of sample both on the PMM (F(2, 43,399) = 147.68, p < 0.001, η² = 0.0068) and MCMC imputed data (F(2, 43,399) = 1,750.56, p < 0.001, η² = 0.074). Post-hoc Tukey’s HSD tests revealed significant differences between all samples in the MCMC imputed data (see Tables 4 and 6, supplementary materials); with sample 2 being lower than sample 1(Mdiff = − 0.437, SE = 0.019, t = − 22.98, p < 0.001) and sample 3 (Mdiff = − 1.185, SE = 0.020, t = − 59.09, p < 0.001). The PMM imputed data showed similar post-hoc Tukey’s HSDs between all pairs of samples, including sample 2 vs. sample 1 (Mdiff = − 0.331, SE = 0.019, t = − 17.16, p < 0.001) and sample 2 vs. sample 3 (Mdiff = − 0.136, SE = 0.020, t = − 6.69, p < 0.001). The model yielded support for H1b, that the onset of the military coup in sample 2 (being associated with perceived societal threat) coincided with lower perceived minority threat and that a decrease in perceived threat in the aftermath of the coup coincided with an increase in perceived minority threat.

Fig. 1 [Images not available. See PDF.]

Interval plot showing the 95% confidence intervals of perceived ethnic minority threat per sample (sample 1 = one year before the coup, sample 2 = one month after the coup and sample 3 = one year after the coup) using complete cases analysis. The dots indicate the average perceived ethnic minority threat per sample.

For perceived Muslim threat (see Fig. 2 below), a Type III one-way ANCOVA was conducted to assess the effect of sample on the dependent variable, controlling for gender, age, education, Buddhist affiliation, and city residence. There was a significant effect of sample in the PMM (F(2, 34,728) = 57.97, p < 0.001, η² = 0.0033) and MCMC (F(2, 34,728) = 108.48, p < 0.001, η² = 0.0062) imputed data (see Table 8 and 10, SM). Post-hoc Tukey’s HSD tests for PMM imputed data revealed significant differences between all samples, with sample 2 being lower than sample 1 (Mdiff = − 0.114, SE = 0.019, t = − 5.91, p < 0.001) and sample 3 (Mdiff = − 0.221, SE = 0.021, t = − 10.52, p < 0.0019, see Table 9, SM). The MCMC imputed data revealed similar patterns as perceived Muslim threat was lower in sample 2 than sample 1 (Mdiff = − 0.215, SE = 0.019, t = − 11.19, p < 0.001) and sample 3 (Mdiff = − 0.272, SE = 0.021, t = − 12.96, p < 0.001, see Table 11, SM). The results yielded support for H1b, that the onset of the military coup (being associated with perceived societal threat) coincided with lower perceived Muslim threat and that a decrease in perceived threat in the aftermath of the coup coincided with an increase in perceived Muslim threat.

Fig. 2 [Images not available. See PDF.]

Interval plot showing the 95% confidence intervals of perceived Muslim threat per sample (sample 1 = one year before the coup, sample 2 = one month after the coup and sample 3 = one year after the coup) using complete cases analysis. The dots indicate the average perceived Muslim threat per sample.

Hypothesis 2

After investigating the between-sample differences on perceived societal threat and perceived minority threat, structural equation models examined hypotheses H2a/b, i.e., that the negative/positive relationship between reported perceived societal threat and support for minority rights will be mediated by higher/lower perceived threat from minorities. While we did have measures of perceived ethnic minority threat and perceived Muslim threat in sample 1, we did not possess measures of perceived societal threat or support for minority rights at that time. Therefore, the structural equation analyses examined H2a/b on data from sample 2 and sample 3 separately. Measurements of perceived ethnic minority threat and Muslim were randomized between participants due to survey length constraints. Therefore, models were estimated separately per mediator and sample.

Perceived ethnic minority threat as mediator

Before estimating the structural equation models, we examined whether missingness in social threat measures and perceived ethnic minority threat is systematically related to key outcomes. Therefore, we conducted logistic regressions predicting missingness in perceived societal threat and perceived ethnic minority threat by their outcomes and control variables. For sample 2, the results indicated that missingness in perceived societal threat was predicted by both greater perceived threat from the majority group (OR = 1.05, p = 0.003, see Table 12, SM) and lower support for minority rights (OR = 0.87, p < 0.001, see Table 14, SM). For sample 3, the results indicated that perceived societal threat was predicted by greater perceived threat from ethnic minorities (OR = 1.13, p < 0.001, see Table 22, SM). These findings suggest that data may not be missing completely at random (MCAR), and multiple imputation was therefore used to minimize potential bias in subsequent analyses³⁹.

First, we examined whether higher perceived ethnic minority threat mediates the positive relationship between perceived societal threat and support for minority rights. Structural equation models (SEM) were specified to examine the relationships between perceived societal threat and support for minority rights through perceived ethnic minority threat, for sample 2 and sample 3 and the PMM and MCMC imputed data separately (see supplementary materials for information on the imputation process). The pattern was consistent across samples and imputation methods, with decreased perceived ethnic minority threat significantly partially mediating the positive relationship between perceived societal threat and support for minority rights (all p < 0.05, see Table 18, 20, 22, 24, SM), with good fit across models (χ²(1) >  = 11.306, p >  = 0.114, RMSEA <  = 0.021, CFI >  = 0.994, TLI >  = 0.925, SRMR <  = 0.004, see Table 19, 21, 23, 25, SM). These analyses therefore offer support for hypothesis 2b, that the positive relationship between reported perceived societal threat and support for minority rights will be mediated by lower perceived threat from minorities.

Perceived Muslim threat as mediator

As for perceived ethnic minority threat, we first tested whether missingness in the independent variables could predict any of the dependent variables of the model. In sample 2, support for minority rights was predicted by missingness on the perceived societal threat measure (OR = 0.86, p < 0.001, see Table 33, SM). In sample 3, perceived Muslim threat was predicted by missingness in the perceived societal threat measure (OR = 1.12, p < 0.001, see Table 42, SM). This motivated us to use multiple imputation on the data used for the structural equation models.

Two structural equation models (SEM) were estimated per sample using PMM and MCMC methods for multiple imputation to handle missing data. The model assessed the direct, indirect, and total effects of predictor variables on the outcome variables adjusted for the control variables, accounting for mediation pathways through perceived Muslim threat. In sample 2, a structural equation model on PMM data indicated a significant total (B = 0.187, SE = 0.029, p < 0.001) and direct effect (B = 0.181, SE = 0.030, p < 0.001), while finding a non-significant indirect effect (B = 0.006, SE = 0.005, p = 0.169, see Table 37, SM). Similarly, results on MCMC imputed data revealed a significant total (B = 0.204, SE = 0.027, p < 0.001) and direct effect (B = 0.202, SE = 0.028, p < 0.001), while finding a non-significant indirect effect (B = 0.002, SE = 0.003, p = 0.512, see Table 39, SM), suggesting that the mediator does not significantly contribute to the overall relationship. There was excellent model fit for both models (χ²(1) >  = 9.546, p >  = 0.179, RMSEA <  = 0.014, CFI >  = 0.997, TLI >  = 0.946, SRMR <  = 0.003, see Table  38, 39, SM). In sample 3, however, structural equation modeling on PMM and MCMC imputed data revealed significant positive total, direct and indirect (all p <  = 0.03, see Table  47, 49, SM) effects, offering support for a partial mediation of the positive effect of perceived societal threat on support for minority rights. Both models had excellent fit (χ²(1) >  = 7.028, p >  = 0.248, RMSEA <  = 0.021, CFI >  = 0.992, TLI >  = 0.946, SRMR <  = 0.002, see Table  48, 50, SM). Therefore, we were only able to collect evidence for hypothesis 2b in sample 3, i.e., that that the positive relationship between reported perceived societal threat and support for minority rights will be mediated by lower perceived threat from minorities.

Discussion

This research investigated the effect of perceived societal threat on perceived threat from minorities. Using data from a three-sample cross-sectional study with a non-WEIRD Myanmarese sample, we tested the effect of the February 2021 military coup and its aftermath on perceived minority threat and support for minority rights. Perceived threat from minorities was found to be higher in sample 1 (collected before the military coup) than sample 2 (collected directly after the military coup). Sample 2 and sample 3 further uniquely aided in observing the relationship between perceived societal threat and perceived minority threat across timepoints (samples). Specifically, perceived societal threat was shown to be higher in sample 2 compared to sample 3. By contrast, perceived threat from minorities was found to be lower in sample 2 compared to sample 3. These results seem to suggest that decreases in perceived societal threat coincide with increases in perceived minority threat (and vice versa). Furthermore, survey responses of perceived societal threat were negatively associated with perceived threat from minorities in sample 2 and sample 3. Finally, the positive relationship between societal threat and support for minority rights was, in most cases, mediated by lower perceived minority threat. Combined, our results provide support for a negative and not a positive relationship between perceived societal threat and perceived minority threat.

However, it is critical to emphasize that we cannot formally substantiate longitudinal claims based on our cross-sectional samples. Our comparisons between samples represent different timepoints, yet the data are not longitudinal, with different individuals assessed at each time point. Thus, while we controlled for several demographic variables, causal claims about the directionality of these effects cannot be made. A longitudinal assessment of the effects of perceived societal threat on perceived threat from minorities would be necessary to validate the temporal dynamics suggested by our findings.

This research adds an important piece of the puzzle to the debate about the relationship between societal threats and threats perceived from minorities. It provides support to the idea that groups are dynamic entities whose focus adapts in response to environmental challenges. Events entirely unrelated to minority groups may shift majority members’ attention away from minority groups, therefore altering the attitudes and behaviors towards minority members. Conversely, our study shows that reductions of perceived societal threat, in turn, relate to increased minority threat. These findings aim to inspire a more comprehensive model of perceived minority threat that includes contextual factors. Moreover, and more importantly, it aims to inspire research that finds pathways to assuage perceived minority threat when no external threat is present. Perhaps creating common goals across minority-majority boundaries may engage the same processes to lessen threat perceptions⁴⁰.

Limitations and future directions

Research in remote non-WEIRD contexts during societal upheaval is particularly challenging and is therefore not without limitations. First, in sample 2 and 3, no data were available on participants’ ethnic background despite measures of perceived ethnic minority threat. This precludes supplementary moderation analyses that could test whether the effects of societal threat observed may vary as a function of the respondents’ own ethnic identity. Future work should measure ethnic group membership to enable such analyses.

Second, while the patterns found are indicative of our second hypothesis, the effect sizes found are overall very small. This may, for example, explain why the indirect path of perceived societal threat on support for minority rights through perceived Muslim threat did not reach significance in sample 2. Third, the study design constraints also allowed for a limited survey length, causing perceived Muslim threat and perceived ethnic minority threat to be single-item measures. Further research should expand these measures to include multiple items.

A fourth limitation relates to certain sample characteristics. Using random domain intercept sampling, we were able to offer our surveys to the entire population of internet users in Myanmar. However, as internet penetration hovers around 35% in Myanmar⁴¹, highly educated males were overrepresented in the survey (see Table 1, SM). Furthermore, although the non-incentivized design facilitated the recruitment of intrinsically motivated participants, it also resulted in a high attrition rate. We employed a rigorous decision approach to deal with missingness, treating missing data appropriate to its nature of missingness, including sensitivity analyses. However, we cannot fully rule out additional sources of bias that may have influenced the missingness in our data⁴².

Finally, an open question remains whether there are moderating variables that predict when societal threat will lead to a reduction versus an increase in perceived minority threat. While our results provide support for a negative relationship between perceived societal threat and perceived minority threat, some literature suggests that the extent to which a perceived threat is shared may determine the nature of the relationship between perceived societal threat and perceived minority threat. Societal threats may be divisive when the minorities are seen as less affected by- or the cause of the threat¹². Conversely, research has shown that perceived threat shared amongst group members can strengthen within-group ties⁴³ and create shared goals to combat the consequences of this threat⁴⁰. Reports on the Myanmar coup have shown that citizens perceived the coup as threatening for the country in its entirety⁵. In the common ingroup identity model, this “sharedness” is exactly what drives common ingroup identification²⁹. While a range of explanations could account for our findings, increased common ingroup identification- a shared threat enhancing perceived belongingess to an overarching group identity- could be a viable candidate. Though we were unable to formally test this mechanism, further research may elucidate whether an increased sense of common ingroup identity may be a mechanism of this negative relationship. Furthermore, this work aims to inspire investigations into the contexts where perceived societal threat may positively relate to perceived minority threat.

In sum, this research demonstrated the interconnectedness of perceived threats from different targets. Perceived threat originating from a certain source may inform threats from a range of other sources. Therefore, our research advocates for a more integrated view on the origins of, the dynamics between and the consequences of perceived threats from different sources to improve more favorable intergroup relations.

Materials and methods

Participants and procedure

This study received IRB approval from The Hebrew University of Jerusalem and the study was performed in accordance with all relevant guidelines and regulations. We collaborated with RIWI (Real-time Interactive World-wide Intelligence; https://riwi.com/), an organization specialized in sampling hard-to-reach populations, to run the studies. Because reaching participants in Myanmar is extremely challenging, our recruitment employed their novel random domain intercept technology. When individuals in Myanmar typed in a website that does not exist, instead of yielding an error (i.e., “this page does not exist”), they were informed about the sampling procedure and invited to a non-incentivized survey⁴⁴. As a result, the survey was randomly distributed amongst internet users in Myanmar (around 35% of the population⁴¹). Due to the broad potential reach (19 milion individuals), the probability of resampling was very minor (i.e., around 0.0002%). Internet users in Myanmar tend to skew male, educated, younger and dwelling in urban centers⁴⁵. As census data tends to be unreliable in this political context⁴⁶, we opted to control- rather than to weigh for these demographic variables in our analyses. None of the demographic variables correlate strongly with the main variables, suggesting no strong influence of their between sample levels (see Table 51, p. 61 SM). Only Myanmarese participants who indicated to be 18 years or older could participate in our study. Informed consent was obtained from all subjects in the study, and the study further complies with all regulations imposed by the IRB of the Hebrew University of Jerusalem.

Three online cross-sectional surveys were carried out. All surveys were translated by local partners of the NGO Digital Public Square that we worked with and were conducted in Myanmarese. Certain items were translated bidirectionally to optimally capture their meaning. The first sample was conducted in March 2020, a year before the military coup occurred. It served as a pre-coup estimate of perceived threat from minorities. The second sample was collected in March 2021, one month after the coup occurred, and provided us with post-estimates of perceived threat from minorities and measures of perceived societal threat and support for minority rights. The third sample was conducted in January 2022, 11 months after the coup, allowing us to test the long-term effects of the coup and providing measures of perceived societal threat, perceived minority threat (i.e., threat from ethnic minorities and Muslims as the most stigmatized religious minority⁴⁷) and support for minority rights.

Length limitations to the survey in sample 1 and sample 2 had some implications for the measurement of the main variables of our study. As indicators of perceived minority threat, in sample 1 and sample 2, participants were either asked about perceived threat from ethnic minorities or Muslims. In sample 3, both were assessed together. In addition, in both sample 1 and sample 2, we were only able to measure perceived ethnic minority threat and perceived Muslim threat with one item for each target group. This item uniquely reflects perceived symbolic threat and not realistic threat. To overcome this limitation, we added extra items in sample 3 referring to perceived realistic ethnic minority and Muslim threat. See supplementary materials p. 32 for analyses using perceived realistic threat items as mediators. We obtained similar results.

Our collaborators of DPS maintain a standard goal sample of 2000 full responses per sample for their surveys. To examine whether this sample size were sufficient to test out hypotheses, we computed the power of this sample size using the mc_power_med app in R Shiny generic power analysis⁴⁸. This analysis showed that, in order to detect a mediation effect with 1 mediator and small correlation effects between all variables (i.e., the main analyses with r = 0.10), a sample of 2000 participants would generate a power of 98% power at the p = 0.05 level, which we deemed as satisfactory. Due to the cross-sectional nature of the samples, comparing between samples is troublesome as sample characteristics may differ. To account for these sample differences, we controlled for participants’ demographics in all our analyses (i.e., gender, age, religion, type of residential community and educational level).

Myanmar’s religious and ethnic landscape is highly complex, with a Bamar-majority adhering to Buddhism alongside diverse ethnic minorities who practice Buddhism, Christianity, Islam, and animist traditions^1,5. In the models examining perceived threat originating from Muslims, all Muslim respondents were excluded, ommitting 480, 610, and 323 participants in samples 1, 2, and 3, respectively. Given the absence of data on ethnic background in samples 2 and 3, we opted to include the entire sample (Bamar and all minority groups) in our models assessing perceived ethnic minority threat, as members of any minority group can conceptualize other ethnic minorities.

Measures

Perceived societal threat

Perceived societal threat as a result from the coup was assessed solely in the second and third sample by asking participants whether they perceived each of the listed threats related to the military coup. Therefore, while we consider an increase in perceived societal threat from sample 1 to sample 2 as very plausible, we cannot test this formally. The measures were inspired by literature on symbolic and realistic threats⁴⁹ and the same items were used in sample 2 and sample 3. The items were formulated generally, without referring to specific events in order to maximize relevance and generalizability across samples. We asked, “which of the below domains do you feel are threatened by the military takeover?”. Response options were “the rights and freedoms of the entire population”, "the rights and freedoms of minorities”, “Your personal security and safety”, “The security and safety of your community” and “your personal financial safety”. The individual perceived societal threat scores were a sum score of the number of items the participant agreed with. A confirmatory factor analysis suggested a one factor model encompassing the perceived societal threat as plausible in both sample 2 (CFI = 0.98, RMSEA = 0.05, SRMR = 0.02) and sample 3 (CFI = 0.99, RMSEA = 0.01, SRMR = 0.01).

Perceived ethnic minority threat

Perceived ethnic minority threat was assessed in all 3 samples with one item, again inspired by literature on symbolic and realistic threat⁴⁹. We asked “Do you agree or disagree? Ethnic minorities are a threat to Myanmar culture and traditions”. Participants then responded using a 1 = strongly disagree to 5 = strongly agree Likert scale.

Perceived Muslim threat

Perceived Muslim threat was assessed in all 3 samples with one item, again inspired by literature on symbolic and realistic threat⁴⁹. We asked “Do you agree or disagree? Muslims are a threat to Myanmar culture and traditions”. Participants then responded using a 1 = strongly disagree to 5 = strongly agree Likert scale.

Support for minority rights

Support for minority rights was only assessed in the second and third sample with 1 item. Participants answered “To what degree do you value the below? The right to protection for all minorities.” on a 1 = not at all to 5 = to a very large extent Likert scale.

Demographics

In all samples, participants answered a brief demographic questionnaire in which their gender, age, education, religion, and type of community of residence (village, city or town) were assessed. We controlled for all demographic variables in all our analyses (see Table 1). For certain analyses, convergence issues were encountered when including all religious denominations, particularly for the multiple imputation procedures, as certain religious denominations were absent in certain samples. To ensure consistency across the analyses, we concentrated on the primary religious distinction in Myanmar society, namely Buddhist (majority) versus Non-Buddhist (minority) groups.

Table 1. Descriptives on the missingness across the three samples.

Sample	NA support minority rights	NA perceived minority threat	NA perceived Muslim threat	NA societal threat	Total N rows
Sample1	12,164 (not measured)	9510 (78.18%)	9663 (79.44%)	12,164 (not measured)	12,164
Sample2	19,909 (91.57%)	16,560 (76.17%)	16,777 (77.17%)	18,200 (83.71%)	21,741
Sample3	6967 (73.32%)	6908 (72.70%)	6857 (72.16%)	7322 (77.06%)	9502

Missing data treatment

Our random domain interception method allowed us to attract a sizeable number of individuals to our study. However, due to the non-incentivized nature of our sampling procedure (i.e., incentives are ethically problematic in these sensitive contexts), we were confronted with considerable non-response rates amongst our studied variables. Our missing data treatment followed three steps: (1) mapping out the percentage of missing data per sample, (2) assessing the systematicness of missingness and (3) a description of our handling of missingness in the data.

First, see below for a table with the non-response rate per measure per sample for the main variables. The proportion of missing data ranged from 73 to 92% per variable.

Second, we assessed the systematism and nature of the missingness. Missing data can be classified into three types: Missing Completely at Random (MCAR), where missingness is unrelated to any variable; Missing at Random (MAR), where missingness depends on observed variables but not the missing values themselves; and Missing Not at Random (MNAR), where missingness is related to unobserved data (i.e., the missing values themselves or an unobserved variable). Complete Case Analysis (CCA) is appropriate when data is MCAR but may lead to bias if data are MAR or MNAR. Multiple Imputation (MI) is preferred when data is MAR, as it uses observed patterns to estimate missing values, improving accuracy. For MNAR, multiple imputation can also be beneficial, but sensitivity analyses or selection models may be necessary. To test whether the data followed a MCAR pattern, we performed a MCAR test using the little’s MCAR test from the Naniar r-package⁵⁰ on the full dataset. This Hawkins test was significant for the global data structure (p < 0.001), therefore showing evidence against the data being MCAR and warranting use of multiple imputation.

Third, we followed Lee & Charles Huber⁵¹ to handle the missing data in our samples. Multiple imputation may be recommended even at higher proportions of missing data³⁹. A complete records analysis will be unbiased for estimating a correctly specified exposure- outcome relationship if the missingness in any variable in the analysis model is not related to their outcome (given the other variables in the analysis model). Thus, for each main analysis with missingness in the independent variables, we assessed if the missingness of the main predictor variable (missing vs. not missing) is related to the outcome. In case at least one of the outcomes in the model was predicted by missingness in one of the independent variables, we multiply imputed the data.

To generate multiple imputations, we need auxiliary variables. Auxiliary variables are variables that are not part of the main analysis but are correlated with missing values or the probability of missingness. Especially when data are Missing at Random (MAR), including auxiliary variables in multiple imputation (MI) helps reduce bias, improve efficiency, and make the MAR assumption more plausible. These variables can be identified by checking correlations with missingness indicators. Once identified, they should be included in the imputation model, ensuring more accurate imputations.

As it is hard to distinguish between MAR and NMAR missingness, performed sensitivity analyses using two different methods to impute data; Predictive Mean Match (PMM) and Markov Chain Monte Carlo (MCMC). For MAR data, both methods should work evenly well, while for NMAR, MCMC should be preferred³⁹. If we receive the same results using all methods, this should increase our belief that regardless of the missing data pattern, our conclusions hold. Therefore, both methods were used as sensitivity analyses.

Acknowledgements

We acknowledge the indispensable contribution of Digital Public Square in aiding data collection in Myanmar to allow us to conduct our research in this challenging context. This study was funded by a European Research Council [864347] to the fourth author.

Author contributions

Daan Vandermeulen is the main author of the paper, being responsible for formulating the idea of the study, reviewing literature, collecting data, analyzing data and writing up the manuscript. Kinneret Endevelt was responsible for reanalyzing the data and formulating feedback on the whole manuscript. Amit Goldenberg and Eran Halperin contributed by streamlining the idea for the project, giving feedback on data collection, data analysis and co-writing the manuscript.

Data availability

The code and data used for the analyses reported in the main manuscript and the supplementary materials are available at: https://osf.io/93x82/?view_only=73b16bd0e5634733879e35821fccfdf4.

Declarations

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Dittmer, L. Burma or Myanmar? The Struggle for National Identity. (WORLD SCIENTIFIC, 2010). https://doi.org/10.1142/7812.

2. Asia, N. Myanmar coup sparks unprecedented unity of ethnic groups. (2021).

3. Gravers, M; Matthews, B. Exploring ethnic diversity in Burma. J. Soc. Issues Southeast Asia; 2010; 25, pp. 292-294. [DOI: https://dx.doi.org/10.1355/sj25-2h]

4. Parmar, P; Mon, SHH; Beyrer, C. The Rohingya genocide and lessons learned from Myanmar’s spring revolution. Lancet; 2022; 400, pp. 793-795. [DOI: https://dx.doi.org/10.1016/S0140-6736(22)01651-8] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36037826]

5. Thawnghmung, AM; Noah, K. Myanmar’s military coup and the elevation of the minority agenda?. Crit. Asian Stud.; 2021; 53, pp. 297-309. [DOI: https://dx.doi.org/10.1080/14672715.2021.1912620]

6. David, R; Myat, AK; Holliday, I. Can regime change improve ethnic relations? Perception of ethnic minorities after the 2021 coup in Myanmar. Japanese J. Politic. Sci.; 2022; 23, pp. 89-104. [DOI: https://dx.doi.org/10.1017/S146810992200007X]

7. Rothschild, ZK; Landau, MJ; Sullivan, D; Keefer, LA. A dual-motive model of scapegoating: Displacing blame to reduce guilt or increase control. J. Pers. Soc. Psychol.; 2012; 102, pp. 1148-1163. [DOI: https://dx.doi.org/10.1037/a0027413] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/22545745]

8. Jedwab, R; Johnson, ND; Koyama, M. Negative shocks and mass persecutions: evidence from the Black Death. J. Econ. Growth; 2019; 24, pp. 345-395. [DOI: https://dx.doi.org/10.1007/s10887-019-09167-1]

9. Becker, JC; Wagner, U; Christ, O. Consequences of the 2008 financial crisis for intergroup relations: The role of perceived threat and causal attributions. Group Process. Intergroup Relat.; 2011; 14, pp. 871-885. [DOI: https://dx.doi.org/10.1177/1368430211407643]

10. Butz, DA; Yogeeswaran, K. A new threat in the air: Macroeconomic threat increases prejudice against Asian Americans. J Exp Soc Psychol; 2011; 47, pp. 22-27. [DOI: https://dx.doi.org/10.1016/j.jesp.2010.07.014]

11. Krosch, AR; Amodio, DM. Economic scarcity alters the perception of race. Proc Natl Acad Sci U S A; 2014; 111, pp. 9079-9084.2014PNAS.111.9079K1:CAS:528:DC%2BC2cXpsVamu70%3D [DOI: https://dx.doi.org/10.1073/pnas.1404448111] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/24927595][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4078865]

12. Adler, E; Hebel-Sela, S; Leshem, OA; Levy, J; Halperin, E. A social virus: Intergroup dehumanization and unwillingness to aid amidst COVID-19 − Who are the main targets?. Int. J. Intercult. Relat.; 2022; 86, pp. 109-121. [DOI: https://dx.doi.org/10.1016/j.ijintrel.2021.11.006] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34866714]

13. Fuochi, G., Boin, J., Voci, A. & Hewstone, M. COVID-19 threat and perceptions of common belonging with outgroups: The roles of prejudice-related individual differences and intergroup contact. Pers Individ Dif175, (2021).

14. Jedwab, R; Khan, AM; Russ, J; Zaveri, ED. Epidemics, pandemics, and social conflict: Lessons from the past and possible scenarios for COVID-19. World Dev; 2021; 147, 105629. [DOI: https://dx.doi.org/10.1016/j.worlddev.2021.105629] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34866756][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8633882]

15. Brambilla, M; Butz, DA. Intergroup threat and outgroup attitudes : Macro-level symbolic threat increases prejudice against gay men. Soc Psychol; 2013; 44, pp. 311-319. [DOI: https://dx.doi.org/10.1027/1864-9335/a000127]

16. Castano, E. In case of death, cling to the ingroup. Eur J Soc Psychol; 2004; 34, pp. 375-384. [DOI: https://dx.doi.org/10.1002/ejsp.211]

17. Greenberg, J; Kosloff, S. Terror management theory: Implications for understanding prejudice, stereotyping, intergroup conflict, and political attitudes. Soc Personal Psychol Compass; 2008; 2, pp. 1881-1894. [DOI: https://dx.doi.org/10.1111/j.1751-9004.2008.00144.x]

18. Jost, J. T., Gaucher, D. & Stern, C. ‘The world isn’t fair’: A system justification perspective on social stratification and inequality. in APA handbook of personality and social psychology, Volume 2: Group processes. 317–340 (American Psychological Association, Washington, 2015). https://doi.org/10.1037/14342-012.

19. Choma, BL; Jagayat, A; Hodson, G; Turner, R. Prejudice in the wake of terrorism: The role of temporal distance, ideology, and intergroup emotions. Pers Individ Dif; 2018; 123, pp. 65-75. [DOI: https://dx.doi.org/10.1016/j.paid.2017.11.002]

20. Doosje, B., Zimmermann, A., Küpper, B., Zick, A. & Meertens, R. Terrorist Threat and Perceived Islamic Support for Terrorist Attacks as Predictors of Personal and Institutional Out-Group Discrimination and Support for Anti-Immigration Policies - Evidence from 9 European Countries. Revue Internationale de Psychologie Sociale22 (2009).

21. Zunes, S. Europe’s refugee crisis, terrorism, and islamophobia. Peace Rev.; 2017; 29, pp. 1-6. [DOI: https://dx.doi.org/10.1080/10402659.2017.1272275]

22. Verkuyten, M; Kollar, R; Gale, J; Yogeeswaran, K. Right-wing political orientation, national identification and the acceptance of immigrants and minorities. Pers Individ Dif; 2022; 184, 111217. [DOI: https://dx.doi.org/10.1016/j.paid.2021.111217]

23. Gump, BB; Kulik, JA. Stress, affiliation, and emotional contagion. J Pers Soc Psychol; 1997; 72, pp. 305-319.1:STN:280:DyaK2s3mtV2htA%3D%3D [DOI: https://dx.doi.org/10.1037/0022-3514.72.2.305] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/9107002]

24. Drury, J; Cocking, C; Reicher, S. Everyone for themselves? A comparative study of crowd solidarity among emergency survivors. Br. J. Soc. Psychol.; 2009; 48, pp. 487-506. [DOI: https://dx.doi.org/10.1348/014466608X357893] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18789185]

25. Andrighetto, L; Mari, S; Volpato, C; Behluli, B. Reducing competitive victimhood in Kosovo: The role of extended contact and common ingroup identity. Polit Psychol; 2012; 33, pp. 513-529. [DOI: https://dx.doi.org/10.1111/j.1467-9221.2012.00887.x]

26. Dezecache, G. Human collective reactions to threat. Wiley Interdiscip Rev Cogn Sci; 2015; 6, pp. 209-219. [DOI: https://dx.doi.org/10.1002/wcs.1344] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26263225]

27. Garcia, D; Rimé, B. Collective emotions and social resilience in the digital traces after a terrorist attack. Psychol Sci; 2019; 30, pp. 617-628. [DOI: https://dx.doi.org/10.1177/0956797619831964] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/30865565]

28. Myrick, R. Do external threats unite or divide? Security crises, rivalries, and polarization in American foreign policy. Int Organ; 2021; 75, pp. 921-958. [DOI: https://dx.doi.org/10.1017/S0020818321000175]

29. Gaertner, S. L. & Dovidio, J. F. Reducing intergroup bias: The common ingroup identity model. (Psychology Press, 2000).

30. Blanchard, AL; Caudill, LE; Walker, LS. Developing an entitativity measure and distinguishing it from antecedents and outcomes within online and face-to-face groups. Group Process. Intergroup Relat.; 2020; 23, pp. 91-108. [DOI: https://dx.doi.org/10.1177/1368430217743577]

31. Moskalenko, S; McCauley, C; Rozin, P. Group identification under conditions of threat: College students’ attachment to country, family, ethnicity, religion, and university before and after September 11, 2001. Polit Psychol; 2006; 27, pp. 77-97. [DOI: https://dx.doi.org/10.1111/j.1467-9221.2006.00450.x]

32. Porat, R., Tamir, M., Wohl, M. J. A., Gur, T. & Halperin, E. Motivated emotion and the rally around the flag effect: Liberals are motivated to feel collective angst (like conservatives) when faced with existential threat. Cogn Emot0, 1–12 (2018).

33. Riek, BM; Mania, EW; Gaertner, SL; McDonald, SA; Lamoreaux, MJ. Does a common ingroup identity reduce intergroup threat?. Group Process. Intergroup Relat.; 2010; 13, pp. 403-423. [DOI: https://dx.doi.org/10.1177/1368430209346701]

34. Horsey, R. One year on from the Myanmar Coup. International crisis group (2022).

35. Min Zaw, O. & Tonnesson, S. Counting Myanmar’s Dead: Reported civilian Casualties since the 2021 Military Coup, PRIO Paper. (2023).

36. Bar-Tal, D., Abutbul-Selinger, G. & Raviv, A. The Culture of Conflict and Its Routinisation. in The Palgrave Handbook of Global Political Psychology 369–387 (Palgrave Macmillan UK, London, 2014). https://doi.org/10.1007/978-1-137-29118-9_21.

37. Rubin, D. B. (1987). Multiple Imputation for Nonresponse in Surveys. Wiley. https://doi.org/10.1002/9780470316696

38. Buuren, S. van, & Groothuis-Oudshoorn, K. (2011). Mice: Multivariate imputation by chained equations in R. J Stat Softw45(3). https://doi.org/10.18637/jss.v045.i03

39. Lee, KJ et al. Framework for the treatment and reporting of missing data in observational studies: The treatment and reporting of missing data in observational studies framework. J Clin Epidemiol; 2021; 134, pp. 79-88. [DOI: https://dx.doi.org/10.1016/j.jclinepi.2021.01.008] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33539930][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8168830]

40. Pettigrew, TF; Tropp, LR; Wagner, U; Christ, O. Recent advances in intergroup contact theory. Int. J. Intercult. Relat.; 2011; 35, pp. 271-280. [DOI: https://dx.doi.org/10.1016/j.ijintrel.2011.03.001]

41. Statista. Internet penetration rate in Myanmar from 2011 to 2020. (2023).

42. Newman, DA. Missing data: Five practical guidelines. Organ Res Methods; 2014; 17, pp. 372-411.2014fdos.book...N [DOI: https://dx.doi.org/10.1177/1094428114548590]

43. Páez, D; Rimé, B; Basabe, N; Wlodarczyk, A; Zumeta, L. Psychosocial effects of perceived emotional synchrony in collective gatherings. J Pers Soc Psychol; 2015; 108, pp. 711-729. [DOI: https://dx.doi.org/10.1037/pspi0000014] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/25822033]

44. Seeman, N. Funding IVF in Quebec: Mining the web to assess public support for policy change. ElectronicHealthcare; 2011; 9, pp. e3-e6.

45. Thaw Tar Min, Fife, E. & Bohlin, E. Consumer demand for the mobile Internet in a greenfield emerging market: The case of Myanmar. In 20th Biennial Conference of the International Telecommunications Society (ITS): ‘The Net and the Internet Emerging Markets and Policies’ , Rio de Janeiro, Brazil, (International Telecommunications Society (ITS), Calgary, 2014).

46. Oh, S.-A. Counting and Being Counted: Ethnicity and Politics in Myanmar’s Census. (2014).

47. Fagan, A. Minority rights in Myanmar: Negotiating identity politics and human rights. In The Noble Banner of Human Rights 240–275 (Brill Nijhoff, 2018). https://doi.org/10.1163/9789004376960_007.

48. Schoemann, AM; Boulton, AJ; Short, SD. Determining power and sample size for simple and complex mediation models. Soc Psychol Personal Sci; 2017; 8, pp. 379-386. [DOI: https://dx.doi.org/10.1177/1948550617715068]

49. Stephan, W. G. & Stephan, C. W. An integrated threat theory of prejudice. in The Claremont Symposium on Applied Social Psychology: Reducing prejudice and discrimination (ed. S. Oskamp) 23–45 (Lawrence Erlbaum Associates Publishers, 2000).

50. Tierney, N. & Cook, D. Expanding Tidy data principles to facilitate missing data exploration, visualization and assessment of imputations. J Stat Softw105, (2023).

51. Lee, J. H. & Charles Huber, J. Evaluation of Multiple Imputation with Large Proportions of Missing Data: How Much Is Too Much? Iran J Public Health50https://creativecommons.org/licenses/by-nc/4.0/ (2021).