Discrimination, Inclusion, and Anti-System Attitudes among Muslims in Germany

Discrimination and Anti-system Attitudes

Research on the political attitudes of Muslims in the West has generally taken two approaches. The first is to argue that Muslims in the West are likely to be relatively more liberal, secular, and prodemocracy than their counterparts who have not lived in the West. Migration to the West has been thought to strengthen democratic norms among migrants and in turn to help facilitate the global spread of democracy (Atkinson, 2010; Docquier et al., 2016; Spilimbergo, 2009). Migrants who live in the West generally develop more positive attitudes about the host country and, in a process of acculturation, often adopt its norms and values, becoming more supportive of democracy, more tolerant, and more politically active (Chauvet and Mercier 2014; Dana 2018; Pérez-Armendáriz and Crow 2010). In a similar vein, Grewal (2020) finds that Islamists who lived in secular democracies became more supportive of secularism and less intent on implementing sharia.

On the other hand, there are also notable exceptions of Muslim migrants who became more antagonistic toward the host country and its values. Sayyid Qutb, an Egyptian poet and literary critic, became so disenchanted by his 1948–50 stint in the United States that he joined the Muslim Brotherhood and eventually wrote tracts that inspired jihadist groups like Al-Qaeda (Calvert 2009; Haddad 1983). More recently, over 6,000 Muslims from Western Europe and North America joined the Islamic State in Iraq and Syria (ISIS) (Barrett 2017).

Why do some Muslim migrants become more democratic and secular, while others, albeit a minority, embrace violence and Islamism? We argue that of fundamental importance is the lived experience of migrants abroad. Among the mechanisms driving acculturation and norm diffusion is that a migrant's personal situation improves in the host country (Careja and Emmenegger 2012; Chauvet, Gubert, and Mesple-Somps 2016). Crediting their new environment for this improvement, they are more likely to adopt its norms and values.

But many migrants do not encounter an improvement, but rather discrimination and hostility by their host communities and governments. Muslims in the West, for instance, face discrimination in the labor market, hate crimes in their neighborhood, negative and stereotypical portrayals in the media, and Islamophobic policies by local and national governments (Awan and Zempi 2017; Bleich and van der Veen, 2021; Lajevardi 2020). Such discrimination has been found to increase the salience of in-/outgroup identities (Haddad 2007) and to breed hostility toward the West (Kteily and Bruneau 2017; Mitts 2019). Others find that discrimination reduces trust in government and support for democracy (Babhoutak, Kavadias, and Vicente 2020; Bilodeau 2017; Maxwell 2010).

We therefore argue that discrimination should serve as a moderator for whether migrants come to embrace or reject the host country and its values. Migrants who face discrimination and hostility are likely to reject the system that produced such prejudice. In the secular democracies of Europe and North America, discrimination should produce lower support for both democracy and secularism.2 In rejecting the West, discrimination should on average also produce more support for Islamism and for violence in defense of Islam. For simplicity, we will refer to these four attitudes collectively as “anti-system attitudes,” although we acknowledge there are major differences between them.

The example of Sayyid Qutb is again illustrative. One reason for Qutb's disenchantment with the West and his turn to Islamism was his experience with racism in the United States. As a dark-skinned man in Greeley, Colorado, Qutb encountered segregated parks and lunch counters and was denied entry into a movie theater (Brogan 2003; Calvert 2009). “When I was in the United States,” Qutb (2020, pp. 122–3) would later write, “I saw them treating coloured people, especially those who were Muslim, with hateful arrogance and much brutality.” Part of his critique of the West in turn became its “evil and fanatic racial discrimination” (quoted in Klein 2005). “The White man, whether European or American, is our first enemy” (Qutb, 1952, p. 1217).

But there is also reason to believe that these effects might vary by the type of discrimination faced. Among many typologies,3 one important distinction is between personal discrimination (whether you have been personally affected) and group discrimination (whether you believe your group faces discrimination). Psychologists have demonstrated that personal discrimination tends to have a greater impact on individual well-being, consistently correlating with lower self-esteem (Schmitt et al. 2014; Verkuyten 1998). Perceptions of group discrimination, by contrast, occasionally correlate with higher self-esteem (Bourguignon et al. 2006; Hagiwara, Alderson, and Mezuk 2016; Stevens and Thijs 2018). Group discrimination increases identification with the ingroup (Tajfel and Turner 1986), which can serve as a buffer increasing individual well-being (Branscombe, Schmitt, and Harvey 1999).

More recent studies envision an interaction between personal and group discrimination. Individuals who have personally faced discrimination and believe that their group often faces discrimination (i.e., high personal, high group) tend to take the experience less personally and are able to attribute it to an ascriptive category rather than something they as an individual did (Stevens and Thijs 2018; Stroebe et al. 2009). Their strengthened identification with the ingroup reduces the pain of their personal discrimination (Branscombe, Schmitt, and Harvey 1999). They “feel less alone in their plight, thereby alleviating the ill-effects of exclusion” (Bourguignon et al. 2006, 773). Meanwhile, individuals who personally face discrimination but who do not believe that others do too (high personal, low group) tend to take the experience more personally and are more likely to feel that they were at fault in some way. Armenta and Hunt (2009) therefore find that individuals perceiving high personal but low group discrimination have the lowest self-esteem.

We extend these psychology studies on self-esteem to the political attitudes under investigation here. We argue that Muslims perceiving both personal and group discrimination might find solace in their shared experiences, and not be so negatively affected as to develop anti-system attitudes. By contrast, Muslims who believe that they suffer alone likely take the experience more personally and become more affected by their discrimination. In turn, the anti-system attitudes that we believe result from discrimination should be especially strong among this group that thinks they suffer alone. They should exhibit the strongest rejection of the system.

It is worth noting that there is not one particular demographic profile or socioeconomic group that is most likely to feel this way. This feeling of suffering alone cuts across demographic groups. It could be the poor, inner-city youth with a criminal record who believes the police uniquely single them out. It could be the older, educated, member of the Muslim Brotherhood with a beard or a veil who thinks their visible religiosity and political affiliation gives them extra attention from the authorities. Or it could be the wealthy, well-integrated citizen who has married a native German and thus has more interactions with and exposure to everyday discrimination. What unites these diverse profiles is this belief that they are uniquely discriminated, more so than other Muslims. It is this group, we argue, that are likely to be most affected by their discrimination and in turn to embrace anti-system attitudes.

On its face, this hypothesis is somewhat counterintuitive. After all, individuals who believe that both they and their group face discrimination likely have strong ingroup identification, which could in theory contribute to hostility toward the outgroup and their associated values. However, we contend that what is most important for individual attitudes is the psychological toll of discrimination. Those who suffered the most from prejudice, even if they do not have the strongest ingroup identification, should see the largest effects on their political attitudes. Figure 1 summarizes our contribution relative to this traditional assumption in the literature.

Survey of Muslims in Germany

To test these hypotheses, we conducted a telephone survey of 1,330 Muslims in Germany. Germany hosts the second-largest Muslim population in Europe (5 million) and has admitted the highest number of Syrian refugees (530,000). Unlike other European leaders, former German Chancellor Angela Merkel sought to welcome and include Muslims, arguing that “Islam belongs to Germany” (Rinke 2015) and that “we are stronger together” (Osborne 2017). German Muslims generally integrate well and express high trust in government (Doerschler and Jackson 2012). Germany is thus one of the few cases where we can develop a credible priming experiment (see below) to highlight inclusive rhetoric and policies.

However, the past decade has also been marked by growing anti-Muslim sentiment and hostility. In 2016, during the height of the “refugee crisis,” only 29% of native Germans had an unfavorable view of Muslims in their country, and only 31% viewed the influx of Iraqi and Syrian refugees as a major threat to Germany (Pew Research Center 2016). By 2019, however, 52% of native Germans viewed Islam as a threat (El-Menouar 2019), and 44% saw it as “fundamentally incompatible with German culture and values” (Sahgal and Mohamed 2019). Hate crimes against refugees spiked 18-fold between 2013 and 2015 (Frey 2020, p. 684) and totaled 9,400 between 2015 and 2019 (Riaz, Bischof, and Wagner 2021, p. 1). German police created a new category of Islamophobic crimes in 2017, documenting nearly 1,000 each year since (The Local 2021). Anti-Muslim sentiment has also helped fuel the rise of the far-right Alternative for Germany (AfD) party (Arzheimer and Berning 2019; Doerr 2021), which has seen its vote share increase from 4.7% in 2013 to 12.6% in 2017. Germany is thus not immune from the wave of Islamophobia sweeping across Europe, making it a useful case for our study on Muslims’ perceptions of discrimination.

Our survey of 1,330 Muslims in Germany was conducted by AMR-Advanced Market Research between April 8 and May 12, 2020 using a computer-assisted telephone interview (CATI) method. Building off the EURISLAM surveys (Koopmans 2016), the sampling frame was created through random digit dialing (both landline and cellphone), linking those numbers to names through phone directories and social media profiles, then employing an onomastic screening identifying typical Turkish, Arab, Pakistani, and Ex-Yugoslav names. To minimize fears of surveillance (Barreto and Dana 2019), AMR was transparent that they were conducting an academic survey on behalf of the Brookings Institution, and they did not describe the survey as targeting only Muslims. The response rate was 24%, slightly higher than for EURISLAM's Germany sample. Each survey question, including anti-system attitudes, likewise had high response rates (about 95%, see the online appendix, pp. 3–4), suggesting respondents were not skeptical or distrustful of the survey. The survey was conducted in German, Turkish, or Arabic, depending on the respondent's selection (96% chose German).

To gauge the representativeness of the survey sample, Table 2 (the online appendix, p. 2) compares it to the German Federal Office for Migration and Refugees (BAMF, 2020) survey of 5,200 Muslims. The comparison reveals a fairly representative sample on most demographics. The sample was about 50–50 men/women, with an average age of 32.7 years. Respondents had lived an average of 24 years in Germany, and 66% were German citizens. About 44% were of Turkish origins, with another 20% from the Middle East (including 10% from Syria). The vast majority lived in West Germany (87%) or Berlin (8%, see the online appendix, p. 2, Figure 8). Most self-described as Sunni (60%) or as “just a Muslim” (18%), with another 11% as Shia, 7% as Alevi, and 1.4% Ahmadi. Nearly 7% had married a non-Muslim, and about 30% had a university education.

Compared to the Federal Office for Migration and Refugees (BAMF) (2020) survey, our sample slightly overrepresents German citizens and North African Muslims and underrepresents first-generation and Southern European Muslims. Given that our intent in this article is to compare attitudes within the sample, and not to extrapolate from it to the population, we err on the side of simplicity and leave the sample unweighted for the main analyses, controlling for these variables instead. Moreover, weights are not yet permitted in the parallel-design causal-mediation analysis. However, we show in the online appendix that all other results are robust to weighting the survey to the BAMF (2020) population estimates (see p. 13, Tables 9 and 10).

Personal Discrimination

We begin with our independent variable, perceptions of personal discrimination. The survey asked: “Have you personally faced discrimination or hostility in Germany?” Despite Merkel's efforts, the majority of respondents—53%—said they have experienced discrimination or hostility (Figure 2, left). To those responding “yes,” the survey followed up with a question asking what sort of discrimination they have faced (“check all that apply”). Figure 2 (right) shows that most reported “verbal harassment” or “discrimination in the workplace or academic institutions,” with smaller numbers reporting physical violence, property destruction, and banning of religious apparel.

While our article focuses on the effects of personal discrimination, it is worth briefly exploring its causes. Figure 3 (left) plots the coefficients from a multivariate linear regression model predicting personal discrimination. Beyond demographics, the model also includes (but the figure does not show) our priming and natural experiments, which we describe later on.

The results show that perceptions of personal discrimination are more prevalent among four demographics. First, Muslims who have become German citizens, as well as those who have married non-Muslims, tend to perceive more personal discrimination. These counterintuitive findings are in line with recent literature suggesting that better integrated Muslims perceive more discrimination, whether due to higher expectations of equal treatment or due to a greater ability to detect discrimination (Lajevardi et al. 2020; Yazdiha 2019). Additionally, younger Muslims, and supporters of the Green party, perceive greater discrimination.

To lend this question greater external validity, we also explore how factors outside of the survey shape these perceptions of discrimination. In particular, Figure 3 (right) shows that spatially, incidents of hate crimes against refugees in a respondent's state (from Bencek and Strasheim, 2016) also correlate with greater perceptions of personal discrimination. That suggests that the perceptions of discrimination captured in our survey likely reflect visceral experiences stemming from real-world events.

Anti-System Attitudes

We hypothesized that personal discrimination will produce higher anti-system attitudes. We measure four such attitudes (see histograms in the online appendix, p. 3). First, we measure support for violence in defense of Islam. Respondents were asked whether “violence against civilian targets is often, sometimes, rarely, or never justified to defend Islam from its enemies?” Support for violence was low: only 18% said it was sometimes or often justified. Second, we measure support for sharia. We asked a yes/no question: “If Muslims one day became a majority in your city, would you favor having local sharia courts for Muslims?” Implementing sharia is a central objective for most Islamist groups, who seek to Islamize their society (Grewal et al. 2019; Hamid 2016). Only 22% favored establishing local sharia courts.

We also examine two pro-system sentiments: support for democracy and support for secularism. We first asked whether “despite its flaws, democracy is the best form of government,” and second, whether “religion and politics should be separate.” Both were asked on a 4-point scale from “strongly disagree” to “strongly agree.” Overall, 78% of respondents agreed or strongly agreed that democracy is the best form of government, and 82% agreed or strongly agreed that religion and politics should be separate.

In line with Hypothesis 1, bivariate correlations (Figure 4) suggest that personal discrimination indeed correlates with all of these four attitudes. First, Muslims who have been personally discriminated were significantly (p < 0.001) more supportive of violence in defense of Islam. The effect size is substantively quite large, doubling the percent saying violence is sometimes or often justified (12% to 25%). Likewise, Muslims who have personally faced discrimination were significantly (p < 0.001) more supportive of local sharia courts, with the effect size more than doubling from 12% to 31%.

Similar patterns emerge for the two pro-system sentiments. Muslims who have experienced discrimination were significantly (p = .026) less supportive of democracy, with support dropping from 83% to 78%. Finally, personal discrimination likewise correlated with lower support for secularism (p < 0.001), with support dropping from 89% to 79%.

Most of these bivariate correlations hold with the inclusion of control variables (see the online appendix, p. 4, for list of covariates). Tables 4 and 5 (the online appendix, pp. 4–6) present the multivariate results. Personal discrimination continues to display a significant (p < 0.001) correlation with violence, sharia, and secularism, and at similar magnitudes as in the bivariate correlation. However, the correlation with democracy loses significance in the presence of controls, although it remains significant when weighting the data to population estimates (the online appendix, p. 13, table 9). Overall, in line with Hypothesis 1, discrimination generally correlates with anti-system attitudes.

While our focus in this article is on these four attitudes, it is worth noting that personal discrimination in fact correlates with every attitude asked in our survey (see the online appendix, pp. 6–7, Figure 10). Muslims who reported personal discrimination were also significantly more likely to say that Muslims will never be accepted in Germany, that the West is trying to destroy Islam, and that they would prefer to leave Germany, among other attitudes.

One potential concern is that because we are “regressing attitudes on attitudes,” it is possible that they are simply capturing the same underlying phenomenon, for instance, a latent antisocial disposition. Although discrimination and anti-system attitudes are very distinct concepts theoretically, we also address this concern empirically in two ways. First, we leverage the aforementioned spatial data on hate crimes to show that these real-world incidents of hostility likewise correlate with higher support for violence and Islamism and lower support for democracy and secularism (see the online appendix, pp. 7–9, Table 6 and Figure 11). Second, we run a factor analysis to examine whether these attitudes can be reduced into a single dimension. On the contrary, we find strong evidence that perceived discrimination and each anti-system attitude are indeed distinct concepts.4

A second potential concern is that the question about personal discrimination does not specify whether they perceived discrimination based on their religion or potentially based on their race or foreign origins. We accordingly control for whether they were born abroad (first generation) and for their region of origin. Moreover, given that these perceptions of discrimination correlate with greater support for Islamism and for violence in defense of Islam, they are likely capturing perceptions of discrimination based on religion.

Personal and Group Discrimination

In short, perceptions of personal discrimination appear to correlate with anti-system attitudes. We now move to testing Hypothesis 2: that this relationship should be strongest among Muslims who think they “suffer alone.” That is, an interaction between personal and group discrimination should reveal that Muslims who have personally faced discrimination but who do not think others have too should show the highest anti-system attitudes.

To test this hypothesis, we leverage a survey question capturing group discrimination. We asked: “in general, how often do you think Muslims experience hostility or unfair treatment in Germany?” with answer options of “never, rarely, occasionally, or frequently.” Overall, 61% of respondents said that Muslims occasionally or frequently face hostility (see full histogram in the online appendix, p. 10, Figure 12A). Personal and group discrimination correlate with each other (see the online appendix, p. 7, Figure 10), though the correlation coefficient is not particularly large (0.23). The predictors of perceiving group discrimination (see the online appendix, p. 10, Figure 12B) also tend to differ from perceiving personal discrimination, suggesting that they are measuring different concepts.

Of particular importance for our purposes is the interaction between these two questions: is one's personal experience with discrimination common and shared by others, or is it relatively rare and seemingly unique to them? The plurality of respondents, 38% (N = 508), said that they have personally faced discrimination and that this is a common occurrence (occasional or frequent) for Muslims in Germany. Meanwhile, 15% (N = 200) said they have personally faced discrimination, but that this is not common (never or rare) among Muslims. About 23% (N = 304) said that although they have not personally faced discrimination, it is common among Muslims, while the final group, 24% (N = 318), said they have not personally experienced discrimination nor is it common among Muslims.

The predictors of perceiving personal but not group discrimination confirm that these feelings of “suffering alone” cut across demographic groups (see the online appendix, p. 10, Figure 12C). More recent immigrants are more likely to feel this way but so are the very well-integrated Muslims who have become German citizens and married non-Muslims. Respondents who identify as Muslim Brotherhood are more likely to think they suffer alone, but so do those who identify as secular, relative to those who identify as neither. In short, there is no one demographic profile most likely to feel this way—and indeed, by controlling for these demographics in the analyses, our intent is to show that it is this sentiment, rather than any of these demographic predictors, that matters for anti-system attitudes.

We compute the interaction between personal and group discrimination in two ways. First, we leave the group discrimination variable on its 4-point scale, which maximizes the variation held in the survey question. However, for ease of interpreting interaction effects, we also dichotomize it into those who believe group discrimination is an occasional or frequent occurrence versus one that occurs rarely or never.

Figure 5 presents both results (with accompanying Tables 7 and 8 in the online appendix, pp. 10–12). In line with Hypothesis 2, respondents who think they suffer alone were far more supportive of anti-system attitudes. Among respondents who believe that they personally face discrimination but that Muslims in general do not (high personal, low group), about 50% were supportive of violence, almost three times as many as in the other groups (p < 0.001). Those who think they suffer alone were thus considerably more supportive of violence. Equally strong results obtain for Islamism, where 50% of those who think they suffer alone reported wanting local sharia courts, three times as high as their counterparts (p < 0.001).

Similar patterns emerge for the pro-system attitudes. Only 60% of this group supported democracy, compared to about 80% across the other three groups (p < 0.001). Finally, only 65% of this group supported secularism, compared to about 85% across their counterparts (p < 0.001). Supplementary results show that this group is also the least likely to say “we can learn from the West,” the most likely to want to leave Germany, and the least proud to be Muslim (see the online appendix, p. 12, Figure 13). Those who suffer alone thus seem to face the highest psychological toll and in turn exhibit the highest anti-system attitudes.

Meanwhile, respondents who perceived both personal and group discrimination were no different in any of these attitudes than their peers who had not faced discrimination. These findings are consistent with the psychology studies suggesting that for this group, “the ill-effects of exclusion” are mitigated through increased solidarity with their community (Bourguignon et al. 2006, 773). Indeed, supplementary results show that this group exhibits the highest ingroup/outgroup identities, and in turn threat perceptions (the online appendix, pp. 12–13, Figure 14). And yet, contrary to most political science literature, despite those hardened identities and threat perceptions, they do not also exhibit anti-system tendencies.

In sum, our analysis reveals considerable variation in how Muslims respond to discrimination. In line with Hypothesis 2, the ones most likely to adopt anti-system sentiments are the ones who believe they are uniquely targeted, more so than other Muslims. There is, of course, still additional variation left to explore: while roughly half of those who believe they suffer alone support violence and sharia, half do not. Supplementary results suggest that these feelings might matter more for certain subgroups, the most prominent being the well-integrated.5 Across all four attitudes, feelings of suffering alone did not correlate with anti-system attitudes for noncitizens, but they did so strongly for citizens. Similarly, for three of the four attitudes (sharia, democracy, and secularism), those who identified as secular were also more affected, while for two of the attitudes (violence and sharia), those who married non-Muslims were more affected. All three of these interactions suggest that perceptions of suffering alone might matter most for the well-integrated, who perhaps feel that if they are still being singled out for discrimination despite all they have done to assimilate, there may be no future for them in the West.

Priming Experiment

How might these dynamics be ameliorated? One possibility, exemplified by Chancellor Merkel, is for political leaders to pursue more inclusive rhetoric and policies. Such political inclusion might reduce Muslims’ perceptions of discrimination and in turn reduce their associated anti-system sentiments.

To test this proposition, we develop a between-subjects priming experiment (see Table 1). The experiment was implemented after demographics were collected and prior to all attitudinal questions. In the experiment, the 1,330 respondents were divided into four groups of about 332 respondents each. One group, the control, received no prime, and moved directly to the attitudinal variables. This group will serve as our baseline to assess the effect of each prime. The remaining three groups were primed with factual information about inclusionary/exclusionary rhetoric and policies pursued by prominent German leaders. All information was likely already known by respondents; the goal of the priming experiment was to remind respondents of this inclusion/exclusion and thereby enhance their salience heading into the dependent variables.

The first treatment, Merkel, primed respondents with inclusionary rhetoric by and policies under German Chancellor Angela Merkel, particularly her statement that “Islam belongs to Germany” and factual statements about Germany accepting the largest number of Muslim refugees in Europe and introducing laws to defend Muslims from hate speech (Martin and Hansen 2018). This prime maximizes external validity, allowing us to assess whether Merkel's inclusiveness had any effect on Muslims in Germany. Relative to the control, we hypothesize that the Merkel treatment will reduce feelings of discrimination (mediator) and thereby reduce anti-system attitudes (outcome).

As noted earlier, Merkel's uniquely inclusive rhetoric and policies, relative to other European politicians, makes Germany perhaps the only case for developing such a credible, factual prime. Yet, the cost of leveraging that external validity was limiting its internal validity. In particular, if this Merkel prime produces effects relative to the control, it is difficult to disentangle whether they are necessarily the result of inclusion or simply of some positive association with Merkel. In this regard we were also limited by Germany's political environment. There are few other politicians in government of Merkel's stature who have been as inclusive, in order to develop a parallel inclusionary treatment. Likewise, Merkel's well-known inclusionary rhetoric made an exclusionary Merkel treatment unfeasible; it simply would not be believed. In other words, while the experiment will allow us to conclude that Merkel's inclusionary rhetoric and policies have such and such effect, it will not allow us to conclude definitively whether other politicians might be able to produce similar effects through inclusion.

Given those limitations, we devoted the other primes to instead testing the effect of exclusion, through two relatively more parallel primes. The second treatment group, Seehofer, primed respondents with exclusionary rhetoric from Interior Minister Horst Seehofer, who, like Merkel, hails from the CSU/CDU party and occupies a powerful political position, but who has disagreed with Merkel's pro-Muslim stance. The treatment therefore primes respondents to recall his statement that “Islam does not belong to Germany” (Staudenmaier 2018), as well as factual statements of Germany setting a cap on Muslim refugees6 and the increase in hate crimes. The final treatment group, Gauland, primed respondents with exclusionary rhetoric from AfD leader Alexander Gauland, particularly the same statement that “Islam does not belong to Germany” (Friedman 2017) and the same factual statements from the Seehofer treatment. The two exclusionary treatments are therefore relatively parallel, simply varying whether the actor occupies a role in government or in the opposition. Given that Interior Minister Seehofer can more directly influence national policies, we expected this prime to be stronger.

Figure 6 presents the results (see corresponding Table 11 in the online appendix, p. 15), comparing each prime to the control group. Overall, the Merkel treatment had a number of statistically significant results. First, Merkel's inclusionary rhetoric and policies reduced perceptions of discrimination and hostility. Relative to the control, respondents in the Merkel treatment were significantly (p = 0.039) less likely to say they have personally faced discrimination or hostility (47% vs. 55%). Relatedly, supplementary results in the online appendix (Figure 16, p. 14) show that respondents in the Merkel treatment were marginally (p = 0.09) less likely to say they want to leave Germany (12% vs. 16%) and were significantly (p = 0.005) more likely to say they are proud of being Muslim (87% vs. 79%). While not quite statistically significant, the Merkel treatment also produced slightly lower agreement that the AfD party represents a threat to Muslims (p = 0.13) and lower agreement that the West is trying to destroy Islam (p = 0.11).7

In turn, the Merkel treatment produced lower anti-system sentiments. Relative to the control, the Merkel treatment produced significantly (p = 0.005) lower support for violence (14% vs. 22%) and significantly (p = 0.041) lower support for sharia courts (19% vs. 25%). These are both substantively large effects as well, reducing support for violence and sharia by about one-third each. The prime also produced significantly (p = 0.049) higher support for democracy (84% vs. 78%),8 as well as marginally (p = 0.06) higher agreement that Muslims can learn from Western culture (81% vs. 75%, see the online appendix, p. 14). Curiously, the Merkel treatment did not also increase support for secularism, perhaps because Merkel has been well-known to draw on Christian values in advocating for inclusion (Heneghan 2010; Mueller 2016).

In short, the Merkel treatment resulted in lower perceptions of discrimination and lower support for anti-system attitudes. But is the lower support for anti-system attitudes driven by the lower perceptions of discrimination? To ascertain whether discrimination mediates the effect of the Merkel prime, we develop an innovative, parallel-design causal-mediation analysis. While the priming experiment alone allows us to rule out pretreatment confounders, it leaves open the possibility of posttreatment confounders, an important but often overlooked obstacle to mediation analysis (Imai, Tingley, and Yamamoto 2013). Following this scholarship, we therefore also leverage a second manipulation—a natural experiment—to provide exogenous variation on the mediator. As we discuss in greater detail in the online appendix (pp. 15–18), we timed the survey to co-occur with Ramadan, surveying half of the respondents prior to Ramadan and half during. Building from Colussi, Isphording, and Pestel (2021), we theorize that Ramadan, by increasing the salience of ingroup/outgroup identities, will increase perceptions of discrimination (the mediator).

Figure 7 presents the results of the parallel-design causal-mediation analysis on the three system attitudes shaped by the Merkel treatment: violence, sharia, and democracy. The results show that the effect of the Merkel prime on each of these attitudes was indeed the result of reduced perceptions of discrimination. The average causally mediated effect (ACME), or the effect of the Merkel prime running through perceptions of discrimination, was significant for all three attitudes. Moreover, the average direct effect (ADE) of the Merkel treatment (i.e., any effect not running through discrimination) was not significant; each was a case of perfect mediation. In other words, we can conclude that the (only) reason why the Merkel prime increased support for democracy and reduced support for violence and sharia was because it reduced perceptions of discrimination.

In sum, the data reveal that Merkel's inclusionary rhetoric and policies have a causal effect reducing perceptions of discrimination and in turn reducing support for anti-system attitudes. While we cannot necessarily claim that similar effects would obtain for another politician, the results are at least suggestive that inclusion by political leaders may reduce support for violence and sharia and increase support for democracy.

While the Merkel prime produced effects across the board, the two exclusionary treatments, Seehofer and Gauland, did not.9 One explanation could be that exclusionary sentiments are virtually omnipresent for Muslims in Germany: with respondents already inundated with exclusionary rhetoric, the primes had little additional effect. The fact that the majority of respondents indeed felt personal and group discrimination lends credence to this point. This interpretation also implies an important limit to Merkel's political inclusion: despite her efforts, the baseline rate of discrimination and hostility remain high. One solitary politician, even if they occupy the most powerful political position, cannot fully overcome systemic political and social discrimination.

Discussion and Conclusion

Leveraging an original survey of Muslims in Germany, this study explored the consequences of discrimination on Muslims’ political attitudes. It found that personal experiences of discrimination correlate with anti-system attitudes: higher support for violence and Islamism and lower support for democracy and secularism. Discriminated Muslims appear to reject the system that produced hostility towards them. Moreover, we find that these sentiments are strongest among Muslims who believe they suffer alone—who believe that other Muslims do not also face such discrimination.

The survey also provided unique, causal evidence that political inclusion might mitigate these dynamics. Through priming and natural experiments, it finds that Merkel's inclusionary rhetoric and policies had a causal effect reducing perceptions of discrimination, and in turn reducing these anti-system attitudes. For policymakers concerned about the “radicalization” of Muslim populations, the results underscore that political inclusion may be an effective approach to reducing support for violence and sharia and increasing support for democracy.

It is important to emphasize that our survey results alone can only speak to attitudinal, and not behavioral, outcomes. Our survey, for instance, gauges public support for violence, not actual incidents of it. That said, some subset of supporters do engage in it, and others may provide resources or moral support and hence are important in addressing the broader phenomenon. Moreover, available empirical evidence is consistent with a behavioral effect of inclusion. In 2015–16, when Merkel's inclusionary rhetoric was at its height and the German population as well exhibited low hostility, Germany sent some of the fewest per capita foreign fighters to ISIS (Capatides 2016). Since then, public expressions of support for ISIS have occurred in areas with greater anti-Muslim hostility (Mitts 2019). We show that these ecological correlations are also found at the individual level, with perceptions of discrimination indeed correlating not just with support for violence but with a rejection of the system writ large.

A final contribution of our study is to raise questions about the mechanism thought to link discrimination to violence or other anti-system attitudes. Much of the literature implies that the adoption of violence is a strategic decision that reflects a conscious desire to defend the ingroup from the outgroup. However, Muslims in our survey who perceived both personal and group discrimination, who we confirm see hardened ingroup/outgroup identities and perceptions of threat (p. 12, Figure 14, in the online appendix), were no more supportive of anti-system attitudes (Figure 5). Instead, what appears to be the link between discrimination and anti-system attitudes is the psychological toll of discrimination. Respondents who suffer most from their experience with prejudice—those who believe they suffer alone—were the ones that adopted anti-system attitudes. Building from the psychology literature showing that those who suffer alone harbor the lowest self-esteem, we argue that the psychology of discrimination, rather than identities and threat perceptions, motivate anti-system tendencies.

While beyond the scope of this study, we do not believe the theory is limited to Germany or to Muslims. Any minority group facing discrimination should exhibit the same patterns, with those members feeling uniquely targeted more supportive of anti-system attitudes. At the same time, the relatively more inclusive context of Germany might have limited the theory's generalizability. In more hostile countries where perceptions of group discrimination are near universal, there may not be many who perceive personal but not group discrimination.