Disentangling Societal Discontent and Intergroup Threat: Explaining Actions Towards Refugees and Towards the State

and migration issues. In this research, we seek to disentangle societal discontent and discontent about refugees in order to better understand how each of these plays a role in the actions that people might take on the streets. Our core concern is to assess to what extent each of these predicts pro- and anti-refugee actions, and more direct anti-government action intentions as well. What we aimed to show is that both pro- and anti-refugee actions are fuelled not just by In debates about migration in Western countries, citizens’ concerns about immigrant groups often go hand in hand with concerns about the decline of society as a whole. Societal discontent, however, is a distinct concept and may have its own relations with immigration attitudes, over and above the role of perceived immigrant threat. In a survey of a representative sample of Dutch people (N = 1239), we disentangled societal discontent from intergroup threat with respect to their relationship with different kinds of action intentions regarding refugees (both pro and anti) and intentions regarding the government. Unsur-prisingly, societal discontent predicted support for anti-government protest (which was strikingly high). More importantly, societal discontent independently predicted both pro-refugee and anti-refugee action intentions, over and above intergroup threat. These associations were moderated by intergroup threat: only when refugees were experienced as a threat did discontent predict anti-refugee action intentions. On the other hand, societal discontent predicted more pro-refugee action intentions, but only when people experienced refugees as an enrichment. Thus, despite populist rhetoric, societal discontent is not always tied to anti-immigrant actions. This suggests that refugee sentiments and societal discontent are not exchangeable: societal discontent plays an important role in reactions to immigration.

Across Western democracies, a part of the population experiences a pervasive sense of discontent about society's functioning: distrust of government, pessimism about the future of society and a sense that society is failing the people (Steenvoorden, 2015;van der Bles et al., 2015). In this paper, we study the influence of this global societal discontent on how people respond to the more specific issue of migration of refugees. It is important to unconfound global discontent from specific threats people might experience from refugees for various reasons: in order to better understand what people are demonstrating for (and against), to better understand societal divisions on topics such as migration and thereby to inform practitioners about appropriate handling of such situations.
We focus on migration because on this issue there was recent anecdotal evidence that global and migrationspecific concerns both play a role in demonstrations and other actions, but the influence of each seems difficult to distinguish. In terms of activism, when the EU was confronted with a sudden influx of refugees in 2015, protests against refugees were at the same time directed against failing governments (Ataç et al., 2016). Concurrently, many demonstrations were held to advocate for better treatment of refugees (Boersma et al., 2019). Here, too, at some protests, refugee issues were addressed alongside more global concerns (e.g., neoliberal policies and human rights). This mixing up of global and issue-specific concerns is also evident in terms of argumentation and rhetoric. On the anti-refugee side, populist parties argue that society is in decline and that limiting immigration is necessary to protect it. For example, Donald Trump and Brexit supporters both argued that immigration should be restricted in order to restore society (Edwards, 2018;Goodman & Narang, 2019). On the pro-refugee side, critiques of (neoliberal) contemporary society are accompanied by calls for equality and provision for those in need, including refugees (Boersma et al., 2019). In sum, the motivations behind and the discourse on anti-immigration and pro-immigration movements are based on a mixing up of broader societal discontent and migration issues.
In this research, we seek to disentangle societal discontent and discontent about refugees in order to better understand how each of these plays a role in the actions that people might take on the streets. Our core concern is to assess to what extent each of these predicts pro-and anti-refugee actions, and more direct anti-government action intentions as well. What we aimed to show is that both pro-and anti-refugee actions are fuelled not just by intergroup threats that are specific to refugees but also by more global societal concerns. Anti-government actions, by contrast, are expected to be mainly driven by societal discontent.

Societal discontent and actions against government
We define societal discontent as the feeling or belief that society, at large, is in a state of decline and is poorly functioning. This feeling is rather unspecific, because it is not about a single societal issue, but it is rather about society in general. This latent feeling can manifest itself in several ways and can therefore also be measured in several ways. Indeed, previous research on societal discontent has shown that measures of specific manifestations of societal discontent, such as low political trust (van der Meer & Hakhverdian, 2016), pessimism and unease about the direction that the country is heading (Steenvoorden, 2015), overestimating the prevalence of societal issues (van der Bles et al., 2015), believing that society is in decline (Elchardus & Spruyt, 2016) and considering leadership to be breaking down and social fabric to be eroding (Teymoori et al., 2016), are highly correlated with each other (Elchardus & Spruyt, 2016;Steenvoorden, 2015;Teymoori et al., 2016;van der Bles et al., 2015), suggesting that they are all affected by an underlying general discontent with society at large. Importantly, this societal discontent does not typically originate in discontent with personal circumstances, but rather it originates in discontent with the societal collective (Elchardus & De Keere, 2013;Elchardus & Spruyt, 2016;van der Bles et al., 2015).
To specifically assess the underlying discontent with society at large and to supplement the existing measures, we add another way to measure societal discontent that explicitly phrases discontent in terms of negative sentiment with society at large, by assessing emotions that people experience about society as a whole, such as frustration and dissatisfaction about current society, concern about society's future, and fear that something will go wrong in society.
As the various aspects or manifestations of discontent strongly relate to each other, the different measures of discontent (including very broad measures) also predict similar outcomes, such as voting for parties that are more at the extreme ends of the political spectrum, both leftand right-wing (Steenvoorden & Harteveld, 2018;van der Bles et al., 2017). For example, political distrust is a strong indicator of dissatisfaction with the way society is functioning (van der Meer & Hakhverdian, 2016) and has, in previous research, been associated with voting for populist radical right parties, which seek to radically change societal arrangements concerning immigration and civil rights (Rooduijn, 2017;Rooduijn et al., 2016). Societal pessimism, defined as the belief that society is in decline and that its future looks grim, also strongly predicts radical political voting (Steenvoorden & Harteveld, 2018). A third conceptualisation is that societal discontent is a more generalized perception, a negative Zeitgeist, that citizens in society are collectively suffering the consequences of various societal issues, such as a high prevalence of financial inequality, corruption, crime and asocial behaviour (van der Bles et al., 2015). This negative Zeitgeist has been shown to predict attributing the cause of negative events to society (van der Bles et al., 2015) and voting for parties that protest against the political mainstream in The Netherlands (van der Bles et al., 2017). In the current political landscape in The Netherlands, the parties that protest against the political mainstream usually also protest against migration policies.
Interestingly, both pessimism and negative Zeitgeist are broad measures of discontent and they apply to many aspects of society as a whole, but they still seem efficient in predicting political behavior. In summary, we argue that the differences between all these conceptualisations are limited as they all tap into discontent of citizens with how society, as a whole, is functioning, and all of them predict a desire to (radically) change society. As such, we treat them as measuring the same underlying sentiment.
Prior research has established that societal discontent (most often in the form of lack of political trust) predicts voting at both extremes of the political spectrum, both right and left (Akkerman et al., 2017;Giebler et al., 2020;Hauwaert & Kessel, 2018;Rooduijn, 2017;van der Bles et al., 2017). It therefore seems logical that discontent would also be related to support for radical action against one's own government, such as disrupting public meetings by government officials, harassing politicians or rioting. To our best knowledge, prior research only provides indirect evidence of this. For example, there is work that suggests that a lack of trust in societal structures is associated with uncooperativeness (Tyler, 2006) and that lack of political trust predicts poor compliance with the law (Mariën & Hooghe, 2011). There is also evidence that suggests political distrust is connected to non-radical actions, such as engaging in signing petitions and joining lawful demonstrations (Braun & Hutter, 2014). Indeed, although there is research showing that perceptions of unfairness are related to forms of radical action intentions that are non-normative, or ' outside of the system' (Morales et al., 2020;Tausch et al., 2011), there is no direct evidence that societal discontent relates to willingness to pursue action to radically change or overthrow one's own government and system. The current research addresses this issue.
Importantly, we believe that societal discontent may not just be a factor in the pursuit of systemic change: it is also likely to play a role in people's responses to specific issues, such as the treatment of refugees. As mentioned above, the current paper focuses on refugee issues, because this is one area where the global and specific are often mixed up in terms of discourse as well as actions. As societal discontent is a global perception about society, the specific way in which discontent relates to pro-or anti-refugee actions depends on people's perceptions of what is problematic about refugees and their treatment.

Intergroup threat and pro-or anti-refugee actions
Whether people support or oppose refugees depends on how they perceive refugees and feel about their arrival. Citizens who have concerns about refugees are likely to experience intergroup threat. That is, they fear that refugees can harm them and the society they live in. Intergroup threat theory (Stephan et al., 2009) suggests that threats can be experienced for symbolic and realistic reasons. Symbolic threat stems from a perceived discrepancy in the meaning system (e.g., cultural values, norms and attitudes) of the in-group and out-group. Realistic threat occurs when people perceive the out-group as endangering their health, safety, prosperity or power. Translating this to refugees, they could be perceived as a societal threat because of their cultural influence or for economic reasons (essentially, refugees could be seen as costly).
Threat is a major factor predicting how people respond to an out-group. Threat is associated with prejudice, stereotyping, discrimination, avoidance of the out-group and actions to strengthen the in-group (Stephan & Stephan, 2017) and also political action. For example, citizens experiencing intergroup threat from immigrants are more likely to oppose social policies favouring immigrants (Pereira et al., 2010) and are more willing to engage in collective anti-immigrant behaviour (signing petitions; attending political meetings, see Shepherd, Fasoli, Pereira & Branscombe, 2018). Therefore, we expect that those who feel threatened by refugees are more inclined to take action against refugees, while those who see refugees as unthreatening (or enriching) are more likely to have positive intentions towards refugees.
As mentioned, the threat that refugees represent is often related to discontent with society as a whole. In populist discourse, the two are sometimes confused to such an extent that immigration is considered the key reason for societal decline. In line with this, research shows that experiencing threat is connected to populist voting (Oesch, 2008). Because of this confusion, it is unclear whether the influence of the two can be separated at all-an empirical question this paper hopes to settle. We believe this should be possible as various political groups who do not feel threatened by refugees at all can nevertheless experience strong discontent about society. For example, on the progressive side, those who are discontented about society often mention the inhumane treatment of refugees. Alternatively, it may be that one is extremely discontented with every government action. This means that those who are discontented and do not feel threatened by refugees may take action in favour of refugees, or they may join in actions that signal their discontent with government policy and the state of society. To disentangle how motives are related to different courses of action, therefore, refugee threat needs to be distinguished from discontent about society.
When discontent with society predicts pro-or anti-refugee action So far, using intergroup threat theory and theory on societal discontent, we have argued that people who feel societal discontent are most likely to have action intentions and that discontent with society may predict either more pro-or more anti-refugee action intentions, depending on how one feels about refugees. More specifically, those who feel threatened by refugees are more likely to have anti-refugee action intentions and, especially those who feel discontent about society and feel threatened about refugees, are the ones with the strongest action intentions against refugees. At the same time, we expect that people who do not feel threatened by refugees are likely to have pro-refugee action intentions. More specifically, we expect that among these non-threatened people, especially those who feel discontent about society, are the ones who want to take most action in favour of refugees, as they are unhappy about how society deals with its problems.
Furthermore, we believe it is useful to disentangle actions that differ in their target: actions that are either for or against refugees, and actions that are against the government. By separating the targets, we get a clearer picture of the role of societal discontent. The reason is that we expect that people who feel discontent about society are more likely to have anti-government action intentions, and this can not necessarily be explained by how one feels about refugees. Put another way, anti-government action intentions do not arise out of dissatisfaction with the refugee situation per se but rather with a belief that society is functioning poorly and is in decline: that society needs to change. Thus, regardless of whether people feel threatened by refugees, they may disagree with existing policies, and when they think things go wrong in society, they show stronger anti-government action intentions.

The current research
In the current paper, we disentangle societal discontent from intergroup threat and investigate how both predict various action intentions. We studied this in The Netherlands where, since 2016, there have been protests against refugees, increased support for populist radical-right parties and also a high rate of volunteerism supporting refugees .
If we can empirically distinguish intergroup threat and societal discontent, we expect them to differently predict anti-government action intentions. More specifically, we expected intergroup threat not to be a strong predictor, whilst societal discontent should be positively related to such action intentions (Hypothesis 1).
Further, we expected societal discontent and intergroup threat to independently predict anti-refugee action intentions. We investigated two kinds of negative anti-refugee action intentions: generic action, such as protesting against immigration, and more targeted actions aimed at refugees directly. The latter tap into the intention to correct refugees who behave out of line, considering violence if necessary. We see these actions as more radical anti-refugee actions, but they are also more targeted and personal. With respect to anti-refugee action intentions, we expected that intergroup threat (positively) and societal discontent (positively) would both be predictors (Hypothesis 2a). We also expected that there would be an interaction between the two, such that only among people who feel threatened by refugees, those who experience much societal discontent are more inclined to take action against refugees than those who feel less societal discontent (Hypothesis 2b).
With respect to pro-refugee action intentions, we expected that intergroup threat would negatively predict action (Hypothesis 3a). We also expected an interaction: people who feel more discontent with society are more likely to have pro refugee action intentions than those who feel little discontent, but only when they do not feel threatened by refugees (Hypothesis 3b).

Method Design
This study had a correlational design in which our primary predictor variables were societal discontent and intergroup threat. Our key dependent variables were antirefugee and pro-refugee action intentions and support for action against the government.

Participants
The current study is part of a longitudinal design consisting of four waves (Kuppens et al., 2020). We only discuss the fourth wave here, because only this wave contained all the items regarding action intentions that we were interested in. The sample came from a survey panel from an internet research company. The questionnaire was filled in by 1291 people in 2019 (the response rate was 82%). However, 52 were removed due to bad response quality, and 32 were removed due to incomplete responses, leaving a sample of 1239 people. The sample was univariately representative of the Dutch population based on age (

Power
To estimate whether the sample was large enough to detect a small path coefficient (beta = 0.10) with 80% power (at the alpha < 0.05 level) in the context of structural equation modelling, we conducted Monte Carlo simulations (Wang & Rhemtulla, 2020). Assuming reliable scales (Cronbach's alpha = 0.75), the simulations indicated that our sample size of 1239 was enough to detect small effect sizes with 80% power.

Procedure
Participants were invited by email to participate in the study, conducted in Dutch. They were compensated by the internet research company for participation with tokens they could exchange for discount vouchers. The research was approved by the Ethical Committee of the university. After being given a short introduction on the topic of the questionnaire (refugees) and giving their informed consent for participation, participants filled in scales that assessed their societal discontent, intergroup threat and action intentions. The questionnaire contained more scales that are part of another study and, hence, are not reported here. 1 The full scales used in this study are reported in the Online Supplementary materials (S4).

Societal discontent
Societal discontent was measured using four subscales that each tapped into a subcomponent of societal discontent. Three items were used for the Lack of Trust component (Cronbach's alpha = 0.86, M = 4.03, SD = 1.28) and were answered on a 7-point Likert scale (1 -not at all, 7 -a lot). The items tapped into trust in national government, local government and courts and police (Schneider, 2017). An example item is 'How much do you trust parliament and the government?' (recoded). Societal pessimism Steenvoorden, 2015) was assessed using a single item, 'Which direction is Dutch society going according to you?', which was answered on a 7-point Likert scale (1 -the wrong direction, 4 -stays the same, 7 -the right direction), and recoded to reflect pessimism (M = 4.96, SD = 1.35). Five items were adapted from Van der Bles and colleagues (2017) to indicate Negative Zeitgeist (Cronbach's alpha = 0.88, M = 4.37, SD = 1.09). The items were rated on a 7-point Likert scale (1 -not at all, 7 -a lot). An example items is 'How much does the average Dutch citizen experience: inequality or unfair treatment'. We also assessed Negative Emotions (Cronbach's alpha = 0.92, M = 4.62, SD = 1.33) by creating four 7-point Likert items (1 -not at all, 7 -very much) that reflected negative emotions regarding the current state and future about society. An example item is 'I feel concerned when I think about the future of society'.

Intergroup threat
For this scale, we adapted five items widely used in the literature (Stephan & Stephan, 2017) to reflect realistic and symbolic threat (Cronbach's alpha = 0.91, M = 4.66, SD = 1.48). The five items were answered on a 7-point Likert scale (1 -strongly disagree, 4 -neutral, 7 -strongly agree). Three items were recoded to create a scale in which higher scores meant more threat. An example item is 'The arrival of refugees is generally good for the Dutch economy'.

Action intentions
To disentangle types of actions, we generated items that tapped into intentions that differed in their aim: three items measured anti-refugee actions (e.g., 'To protest against refugees'), two items measured corrective action against refugees (e.g., 'To correct refugees who cross the line, with violence if necessary'), three items measured pro-refugee actions (e.g., 'To demonstrate for rights for refugees') and three items measured support for anti-government actions (e.g., 'The government functions so poorly, that it is best to overthrow the whole system'). Corrective action against refugees was initially measured with three items, but one item was removed from the scale because it was functioning poorly in an exploratory factor analysis (see Results section).
All scales except anti-government actions were rated on a 7-point Likert scale (1 -Absolutely not, 7 -Absolutely), with the prefix 'To what extent would you consider to do one of the following things [action]'. Support for anti-government action intentions were rated on a 7-point Likert scale (1 -Strongly disagree, 4 -Neutral, 7 -Strongly agree).

Factor analysis of action intentions
Because the items for the action intentions did not come from pre-validated scales but were created for the current research purpose, we examined their structure with an exploratory factor analysis (minimum residual, oblimin rotation). This analysis confirmed that the expected 4-factor structure was an appropriate fit to the data (see Table 1). We dropped one item from the corrective action scale, using only two items to measure the construct, because it did not clearly load on one scale.
The final scales were anti-refugee action intentions

Testing the hypotheses
To assess the effect of intergroup threat, societal discontent and their interaction on the four types of action intentions, two structural equation models were fitted. These models specify latent variables of societal discontent and of intergroup threat. In these models, societal discontent was defined by the four distinct indicators of discontent (political trust, pessimism, negative zeitgeist and negative emotions); intergroup threat was defined by the five items of the scale. For the types of action, the respective items from the action intentions scale were used as indicators. The models thus contained six latent variables.
We compared a model with main effects of discontent and threat on the action intentions with a model that additionally included a latent variable interaction between discontent and threat. The reason for two models is because we were interested in main effects as well as interaction effects. By using moderated structural equation modelling instead of regular linear regression with interaction variables, the influence of measurement error in the models was reduced (see Cortina et al., 2019). The chi square test was significant, but that is common in high sample sizes (Bollen & Noble, 2011). Inspecting the residual correlations, none of them were problematic, and the majority (96%) was below 0.10, suggesting limited deviation of the model-implied correlation matrix from the observed correlation matrix.
Intergroup threat and societal discontent were strongly positively covarying predictors, B = 1.24, SE = 0.064, 95%  Table 2 and Figure 1 for an overview of path coefficients).

Model 2: Main effects and interaction effects of societal discontent and intergroup threat
For Model 2, the same latent variables were used as in Model 1, but we included a latent variable to estimate the interaction effect of intergroup threat and societal discontent on the four types of action intentions. To estimate a moderated structural equation model without strict normality assumptions, we followed a well-established procedure (Wu et al., 2013), which builds on the unconstrained approach (Lin et al., 2010). In this procedure, only the most reliable indicators (as determined by the highest standardized loading) of each latent predictor variable are multiplied and used as indicators for the latent interaction variable, as this limits information redundancy and correlated errors, yielding a more parsimonious model. Most importantly, the indicators, their variances and the variance of the latent interaction were not constrained (for the full specification, see Appendix B). All predictor variables were mean-centred prior to computing the interaction variables, and the interaction variables were mean-centred again to prevent spurious correlations with the predictor variables (Lin et al., 2010). To estimate latent means that are on the same metric as the observed variables, we used effect-coding in the mean structure (Little et al., 2006). We estimated this model using maximum likelihood estimation, and therefore report robust standard errors and a Satorra-Bentler scaled test statistic (Coenders et al., 2008).
This model fitted well (χ² (231)   between the observed and model-implied correlations were similar to those in Model 1. Although the explained variance of the action intentions did not differ substantially between models, the model with interaction paths was significantly better than a model with these paths fixed to zero (χ²(4) = 45.58, p < 0.001). Testing the paths individually, only the interaction effect on corrective-refugee action did not significantly improve model fit (χ²(1) = 3.24, p = 0.072).
The (main effect) paths from societal discontent and intergroup threat to the four action intentions were largely the same as in Model 1 (see also Table 2). To better understand the interaction effects, we conducted simple slope analysis (Preacher et al., 2006). The slopes for societal discontent on each action intention were estimated at the average of intergroup threat, one standard deviation below the average (refugees as an enrichment, -1.63) and one standard deviation above the average (refugees as a threat, +1.63). The results are displayed in Figure 2. The effects that are reported below remained significant when controlling for age, gender and education level. The results of those analyses can be found in the supplementary materials (Table S3). Support for anti-government action. Unexpectedly, for support for anti-government action, there was a small interaction effect, B = 0.08, SE = 0.02, 95% CI [0.04; 0.12], p < 0.001, beta = 0.08 (see bottom right panel of Figure 2). This interaction showed that the relation between societal discontent and support for anti-government action intentions was somewhat stronger among those who felt more threatened by refugees, B = 1.29, SE = 0.06, z = 20.78, p < 0.001, than among those who did not feel threatened, B = 1.04, z = 16.17, SE = 0.06, p < 0.001. However, this interaction effect was small, especially compared to the main effect of societal discontent, B = 1.17, SE = 0.05, 95% CI [1.06; 1.27], p < 0.001, beta = 0.80, suggesting in line with hypothesis 1 that indeed on all levels of threat, societal discontent strongly predicts anti-government action intentions.
Anti-refugee actions. For anti-refugee actions, we found support for Hypothesis 2b: there was a significant interaction effect on anti-refugee action intentions, B = 0.13, SE = 0.02, 95% CI [0.09; 0.17], p < 0.001, beta = 0.18 (see top left panel in Figure 2). The slope of societal discontent was not significant when refugees were seen as an enrichment, B = 0.016, SE = 0.046, z = 0.35, p = 0.73, but was significant at the average level of intergroup threat, B = 0.23, SE = 0.041, z = 5.59, p < 0.001, and especially when refugees were seen as more of a threat, B = 0.44, SE = 0.059, z = 7.42, p < 0.001. Taken together, this suggests that societal discontent predicted anti-refugee action intentions, but not when refugees were seen as more of an enrichment.
Refugee-corrective action intentions. We did not find support for hypothesis 2b with respect to refugeecorrective action intentions. Interestingly, simple slopes were significant at the three levels of intergroup threat, and there was no significant interaction effect, B = 0.06, SE = 0.03, 95% CI [0.00; 0.13], p = 0.068, beta = 0.06. Thus, refugee-corrective action intentions were positively predicted by societal discontent, B = 0.52, SE = 0.07, z = 7.04, p < 0.001. Furthermore, quite unexpectedly, even those who did not feel threatened by refugees, but rather saw them as an enrichment, were more likely to have refugee-corrective intentions when they experienced more societal discontent, B = 0.42, SE = 0.09, z = 4.56, p < 0.001. We explored whether this could be explained by a tendency towards authoritarianism by controlling for its main effect and its two-way interaction effect with societal discontent and intergroup threat. Using a 2-item version of an Authoritarian Aggression scale 2 that is commonly used in prejudice research (item 2 and item 4 from (Duckitt et al., 2010), results showed that the positive relation between societal discontent and refugee-corrective action intentions for those low in intergroup threat could be explained by the association between discontent and authoritarian aggression; once authoritarian aggression and its interactions were controlled for, those low in intergroup threat no longer showed a meaningful relation between discontent and refugee-correcting action intentions. Indeed, Authoritarian Aggression functioned as an individual difference variable that predicted these violent behaviour intentions for both people that do and do not feel threat from refugees (see Supplementary, S2, for all parameters), B = 0.28, SE = 0.05, 95% CI [0.18; 0.39], p < 0.001, beta = 0.19. Importantly, the interaction effect of threat and discontent became significant after controlling for Authoritarian Aggression, B = 0.16, SE = 0.07, 95% CI [0.01; 0.30], p < 0.001, beta = 0.15, meaning that for those who did not experience intergroup threat, societal discontent did not predict more intentions to correct refugees, B = 0.18, SE = 0.13, z = 1.41, p = 0.16, while for those who did experience intergroup threat, societal discontent still predicted refugee correcting intentions, B = 0.69, SE = 0.16, z = 4.41, p < 0.001.
Pro-refugee action intentions. As expected, for prorefugee action intentions, there was a significant interaction effect, B = -0.071, SE = 0.02, 95% CI [-0.11; -0.03], p < 0.001, beta = -0.10. The slope at one standard deviation below the mean of intergroup threat was positive and significant, B = 0.30, SE = 0.062, z = 4.85, p < 0.001, and the slope at the mean of intergroup threat was also significant, B = 0.18, SE = 0.047, z = 3.92, p < 0.001, but the slope was not significant at one standard deviation above the mean, B = 0.069, SE = 0.05, z = 1.32, p = 0.19. In line with Hypothesis 3b, we found evidence that societal discontent positively predicted pro-refugee action intentions, but not when refugees were seen as more of a threat.

Discussion
In this research, we sought to disentangle citizens' perceived refugee intergroup threat from societal discontent (the negative feeling that society at large is in decline) with respect to their relation to behaviour towards refugees and the government. More specifically, we examined how they relate to three different kinds of behavioural intentions: negative behavioural intentions towards refugees, positive actions towards refugees, as well as violent action against the government. A study with a Dutch representative sample showed support for the following conclusions.
First, even though societal discontent and intergroup threat are positively related and are strongly connected in populist rhetoric, it is useful to examine their unique relation with behaviour towards refugees and the government. In our research in the Netherlands, and at this time in history, there is a strong positive relation between both variables. Nevertheless, across the models each differentially predicted different kinds of action intentions. Indeed, while societal discontent is positively related to anti-refugee action intentions, it is also related to pro-refugee action intentions: societal discontent and intergroup threat interact, as discontent was positively related to prorefugee action intentions, but only for people who see refugees as an enrichment rather than a threat. This clearly suggests that societal discontent does not always lead to anti-refugee action and can even lead to action to help refugees. The value of distinguishing refugee intergroup threat from societal discontent is further exemplified by the fact that anti-refugee behaviour is predicted by both, and that the two interact in this case as well: discontent was more strongly related to anti-refugee action intentions for those who perceived more intergroup threat. All in all, for both theoretical and empirical reasons, we conclude that distinguishing the relation that each of these variables has with behaviour is both possible and useful.
Second, regarding support for anti-government actions, there were surprisingly high levels of support for anti-government action, especially considering that those items were strongly worded (e.g., items such as 'the government functions so poorly that is best to overthrow the whole system'). The results show they are strongly related to societal discontent and not to refugee threat. Indeed, support for more aggressive actions against the government, and even support for overthrowing the current system with violence, was only predicted by societal discontent. The relationship between societal discontent and support was slightly stronger for those who experienced intergroup threat than for those who did not. Both this finding, as well as the high correlations between societal discontent and both refugee intergroup-threat and intentions, may be attributed to a substantial group in society that desires to act against the government directly because of their dissatisfaction with refugee policies.
Third, we investigated two types of anti-refugee action intentions, and one of them, the refugee-correcting action intentions, yielded unexpected results: that is, those who felt more societal discontent indicated intentions to correct refugees even if they considered refugees more of an enrichment than a threat. This positive relation for those who perceive little intergroup threat could be because corrective behaviour is a compensation for a perceived lack of societal control over certain outcomes and situations. Citizens with societal discontent may perceive that the government is not doing enough and are inclined to take matters into their own hands. An exploratory analysis indicated that authoritarian aggression could explain our unexpected finding. In our sample, societal discontent was connected to a desire for strong rule enforcement in general, and this appears to explain why those who see refugees as an enrichment also showed a positive relation between discontent and refugee-correcting action intentions.

The unique role of societal discontent
Together, our findings suggest that there are very good reasons to study societal discontent more closely: it may be one of the engines of today's revolutionary movements, both on the political right and on the left. The nature that all the actions we examined have in common is that they pursue system change. It makes sense for a generic feeling of systemic failure to underpin such pursuits.
Societal discontent reflects the broad negative feelings and attitudes that many citizens have towards their society, such as a lack of trust in institutions (van der Meer & Hakhverdian, 2016), pessimism about society's future (Steenvoorden, 2015), perceiving leadership and social fabric as eroding (Teymoori et al., 2016), perceiving a high prevalence of various societal issues (van der Bles et al., 2015) and worry and fear about society. This broad and general discontent predicts action intentions, and this broad conceptualization of societal discontent is much more general than attitudes and feelings that have to do with the refugee situation. We contribute to the literature by showing that this general attitude about society has consequences for action intentions regarding quite specific societal issues, such as the refugee situation. This connection between generic discontent and very specific action intentions has several theoretical implications.
Firstly, societal discontent is not just related to specific action intentions towards refugees, neither is it tied to an immigration context. Indeed, there are many more action intentions where societal discontent may play a role. Societal discontent could play a role in every issue that becomes societally contentious, whether it addresses economic policies, the welfare system, education or concerns about banks or any other policy area. In all these potential contexts, we expect societal discontent to fuel intentions for collective action that achieves societal-level change.
Secondly, previous literature found that societal discontent, such as lack of trust, is connected to populist voting (Steenvoorden & Harteveld, 2018). One key contribution of the present research is that it goes beyond voting by examining intentions for regime change, intentions that consider force and violence if deemed necessary. We were surprised about the high level of support for this. Future research is necessary to replicate these findings, primarily because our measures of system change were rather general and did not ask about specific desired changes. An interesting future extension would be to examine the relationship between societal discontent and support for a much broader range of more ideologically inspired actions that achieve certain societal changes. It may be that among those who adhere to a specific ideology, those who experience more societal discontent are more likely to support societal change to achieve this ideology.

Implications for prejudice research
The current research also has implications for prejudice research. First, we observe that the more general relationship of citizens to their society at large has generally not received much attention in prejudice research (for an exception, see the literature on relative deprivation, e.g., Pettigrew et al., 2008). While intergroup threat research emphasizes how out-groups are perceived as threats, it overlooks the broader societal context within which the individual who feels threatened is situated. In our study, societal discontent predicted anti-refugee intentions independent of intergroup threat. Therefore, future research should take societal discontent into account to paint a more complete picture of forms of prejudice and antiimmigrant action.
Secondly, how societal discontent translates into specific behaviour is likely to depend on one's stance towards the issue at stake. Indeed, attitudes that are related to a societal issue, such as intergroup threat in this paper, specify what action direction is chosen, but societal discontent specifies the extent of action intentions. Thus, frustration and worries regarding society at large can translate into action intentions towards society in general or with respect to specific societal issues, such as the refugee situation. Whether these action intentions are pro or anti refugees depends largely on how one relates towards refugees as a group.

Implications for research on populism
With all this in mind, we can comment on the relationship between intergroup threat and societal discontent in political rhetoric and populism research. Although threat and discontent were positively connected, we show that negative feelings about society that are, especially in current Western societies, associated with rightwing populist rhetoric can also be related to (left-wing) pro-refugee action intentions. This notion is exemplified by the change in political circumstances that occurred after Trump got elected in 2016. While during Trump's campaign the conservatives expressed discontent with society, it is likely that societal discontent among liberals mobilized their political (re)actions, such as the Women's March held in the United States in 2017 (Jamieson, 2016). Importantly, extending previous research (e.g., Steenvoorden & Harteveld, 2018), societal discontent seems to be able to predict political behaviour that can be either characterized as more left-wing (such as helping refugees) or right-wing (such as protesting against refugees). This finding that societal discontent can predict pro-social action (in this case, donating money and helping refugees) is in line with research showing that pro-social action can be a form of collective action (Thomas & McGarty, 2018) and can therefore be seen as reflecting a desire to achieve social change. We can conclude that societal discontent is not inextricably linked to intergroup threat, and future research could further investigate the consequences of feeling societal discontent among those who do not feel intergroup threat but rather see specific out-groups such as refugees as an enrichment.

Limitations and directions for future research
First and foremost, the study was cross-sectional in nature, meaning that the causal pathways between societal discontent and intergroup threat and action intentions could not be empirically determined. Nevertheless, drawing upon existing attitude research (Ajzen & Fishbein, 2005), it seems most plausible that threat perceptions and societal discontent precede the action intentions (but also see Olson & Stone, 2005). Future research could further affirm the causal nature.
A second direction for future research is to corroborate the findings in a different context as this study was conducted in The Netherlands. We believe that the current pattern of results will be similar for many Western societies, because similar relationships between, for example, anti-refugee attitudes and societal discontent have been found in other (European) countries (Aschauer & Mayerl, 2019;Rooduijn, 2017). We therefore expect that across European countries, anti-refugee protests will be predicted by societal discontent and refugee intergroup threat. Future replications in other countries are needed to further establish and replicate the findings.
Future research should then also take political orientation into account, which we did not measure in the current study. While we think that all people of all political orientations can experience societal discontent, the response to certain societal issues may be particularly politicized and polarized in certain countries while they are not in other countries, and societal discontent and political orientation may therefore predict different behavioural action strategies.
A limitation of the current research is that the action scales we used were not well-validated, as they were created for the purpose of our study. Although the scales showed good reliability, we do not know their validity and relationship to other concepts. Furthermore, the items for the refugee corrective-action intentions were doublebarrelled. In the way the items were framed, one could agree to the item without necessarily endorsing violence. However, we favour the interpretation that especially those that more strongly agree to the item would also consider violence as an option, and that those who oppose violence, would tend to disagree more. We encourage future research to better assess these violent intentions. Furthermore, we proposed that societal discontent reflects a desire to change society as it currently is. However, we did not investigate this variable, and future research could investigate it as a mediator.
Another limitation is that our scale of government-oriented action intentions did not tap into any constructive positive actions towards the government or society. We showed that discontent predicts support for action intentions against governments, but discontent with society could also theoretically lead to non-aggressive behaviour, such as participation in political campaigns and parties. However, our scale did not assess this constructive path to societal change, and future research is needed to investigate this possibility.
One more issue is that we did include a measure of threat and some authoritarian aggression items in the study, but we ignored other relevant variables, such as Right-Wing Authoritarianism and Social Dominance Orientation (RWA and SDO). These are linked to negative beliefs about out-groups and views about society (Duckitt, 2001). Those who score high on RWA see society as more dangerous and are more likely to see unfamiliar refugees as a threat, and people with high scores on SDO may perceive refugees as a threat to intergroup hierarchies (Duckitt, 2001). Future research could explore how these views of society relate to both threat and societal discontent.
Finally, we do not have evidence with regards to actual behaviour, and the generalization of these findings to actual behaviour is somewhat limited. Although we think the items are valuable and strongly suggest a link with behaviour, future research is needed to establish whether societal discontent in combination with issue-specific attitudes leads to actual behaviours.

Conclusion
Taken together, the findings demonstrate the value of disentangling societal discontent from intergroup threat when trying to understand action intentions related to the refugee debate. The people who are acting towards refugees are doing so for two reasons, which are conceptually quite distinct: the threat posed by refugees or a more general discontent with society as a whole. Although these two attitudes are often confounded (for example in populist rhetoric), discontent with society at large does not always go hand in hand with anti-refugee attitudes and behaviour. People who feel societal discontent can also feel positive towards refugees and express it by engaging in pro-refugee action. Thus, discontent does not always lead to right-wing actions. Instead, societal discontent acts as a negative unspecific feeling or belief that society is in decline, which fuels action intentions concerning a specific societal issue. When societal discontent and anti-refugee attitudes do go hand in hand, it is entirely possible that migrants are used as scapegoats for a much broader spectrum of perceived societal woes.

Data Accessibility Statement
The data used in this study can be accessed at https://osf. io/b3ysu/.

Notes
1 For example, we measured stereotypes and meta-stereotypes regarding the elite, perceived conflict in society, societal misrecognition, authoritarian aggression, feelings towards others in the migration debate, feelings towards refugees and how the country treats refugees, acceptability of ongoing protest behaviours. 2 This scale was part of the questionnaire and measured for another study. It is included here only for exploratory purposes.

Additional Files
The additional files for this article can be found as follows: •