The Nature of Islamophobia: A Test of a Tripartite View in Five Countries.

This article provides an examination of the structure of Islamophobia across cultures. Our novel measure-the Tripartite Islamophobia Scale (TIS)-embeds three theoretically and statistically grounded subcomponents of Islamophobia: anti-Muslim prejudice, anti-Islamic sentiment, and conspiracy beliefs. Across six samples (i.e., India, Poland, Germany, France, and the United States), preregistered analyses corroborated that these three subcomponents are statistically distinct. Measurement invariance analyses indicated full scalar invariance, suggesting that the tripartite understanding of Islamophobia is generalizable across cultural contexts. Furthermore, the subcomponents were partially dissociated in terms of the intergroup emotions they are predicted by as well as the intergroup outcomes they predict (e.g., dehumanization, ethnic persecution). For example, intergroup anger and disgust underpin Islamophobic attitudes, over and above the impact of fear. Finally, our results show that social dominance orientation (SDO) and ingroup identification moderate intergroup emotions and Islamophobia. We address both theoretical implications for the nature of Islamophobia and practical interventions to reduce it.

A growing number of people in the West view Islam and Muslims negatively (e.g., Gallup, 2013). According to a recent survey, 55% of Europe’s population agreed that migration from predominantly Muslim countries should be stopped (Goodwin et al., 2017). Similarly, Muslims are the least favorably viewed, and most dehumanized, religious group in the United States (Kteily et al., 2015; Pew Research Center [PEW], 2017a), with outcomes ranging from institutional discrimination (e.g., “Muslim Ban”) to hate crimes (e.g., Elahi & Khan, 2017). In many parts of the world, Muslims also face arbitrary detention, forced sterilization, torture, and ethno-religious “cleansing” (Human Rights Watch [HRW], 2018; United Nations [UN], 2018).

Yet, despite the prevalence of Islamophobia—commonly used to refer to negative attitudes, emotions, and behaviors toward the Islamic religion and Muslims—no agreed-upon operationalization of the term exists (e.g., Bleich, 2011; Helbling, 2012; Klug, 2012; Uenal, 2016a). This absence is reflected in researchers’ use of disparate items and scales (e.g., Imhoff & Recker, 2012; Lee et al., 2013). Moreover, to the best of our knowledge, these measurement approaches have not been validated cross-culturally. Crucially, due to this dearth of cross-cultural validation, the extent to which Islamophobia’s dimensionality is generalizable across contexts remains unknown (Bleich, 2011; Küntzl, 2008; Uenal, 2016a; but see Kunst et al., 2013 for perceived Islamophobia). Furthermore, Islamophobia’s underlying emotional origins, psychological mechanisms, and its consequences remain understudied (e.g., Choma et al., 2012, 2016; Uenal, 2016a, 2016b, 2017c).

In this paper, we propose, and cross-culturally validate, a tripartite view of Islamophobia, which differentiates between psychologically distinct components. Importantly, we also address a current debate over whether Islamophobia is primarily fear-based or, instead, whether it is better explained as socio-functional threat-based psychological reactions of anger and/or disgust to a social group, in addition to fear (e.g., Ketelaar, 2015; Neuberg & Schaller, 2016; Park et al., 2007; Plutchik, 1980). Finally, we test two individual difference variables (social dominance orientation [SDO] and ingroup identification), which might explain why some individuals react with fear while others display anger and/or disgust in response to Islam and Muslims (e.g., Choma et al., 2016; Matthews & Levin, 2012). Thus, this paper aims to advance research on the nature of Islamophobia by addressing persistent theoretical and methodological issues including Islamophobia’s conceptualization, measurement, cross-cultural validity, emotional origins, and societal consequences.

What Constitutes Islamophobia?

Most commonly, Islamophobia is defined as negative, fear-based attitudes and behaviors toward Islam and its adherents—Muslims (e.g., Bangstad, 2016; Bleich, 2011; Conway, 1997; Halliday, 1999; Klug, 2012). Accordingly, prominent measures of Islamophobia conflate items pertaining to both Islam (religious concept) and Muslims (social group; e.g., Helbling, 2012; Imhoff & Recker, 2012; Lee et al., 2013). Yet, we argue that this comes with several interrelated caveats.

Conflating Individuals and Concepts: Theoretical and Methodological Concerns

In the social psychological literature, stereotyping, prejudice, and discrimination primarily refer to biases toward individuals based on their membership in a group and not toward the cultures, religions, institutions, or ideologies that individuals are a part of (e.g., Allport, 1954; Brown, 2010). In other words, traditional bias measures assess attitudes toward members of social groups (e.g., Jews, Atheists), not the abstract concept the group is adhering to (e.g., Judaism, Atheism). Importantly, research suggests that individuals are evaluated more positively than their larger impersonal representations (sometimes referred to as a person-positivity bias; e.g., Sears, 1983). For example, political party leaders (individuals) are more positively evaluated than the political institution they are part of (i.e., Nilsson & Ekehammar, 1987). Indeed, previous research indicates that Muslims are more positively evaluated than their less personal representations (e.g., Islam; Leibold & Kühnel, 2006). However, whereas modern measures of negativity toward Jews and Judaism differentiate between evaluation of its members (anti-Semitism) and its religion (anti-Judaism); Bilewicz et al., 2013; Bilewicz & Krzeminski, 2010), measures of Islamophobia do not.

Although it may seem reasonable to assume that individuals holding negative sentiments toward the Islamic religion are likely to also be biased against Muslims (Klug, 2012; Miles & Brown, 2003), this relationship remains understudied (Uenal, 2016a). Furthermore, the conflation between Muslims and Islam has also triggered a heated debate about the comparability of Islamophobia with other types of prejudices like anti-Semitism (e.g., Bangstad, 2016; Bleich, 2011; Küntzl, 2008). To address this, here, we explicitly examine both attitudes about Muslims (individuals) and Islam (religious concept). We expect them to be statically distinguishable.

A Secret Islamization of the West: Integrating Islamophobic Conspiracy Beliefs

Another key component of Islamophobia—Islamophobic conspiracy beliefs—also remains understudied and has not yet been integrated into current conceptualizations and measurements (O’Donnell, 2018; Swami et al., 2018; Uenal, 2016c). The belief in an Islamic conspiracy or a secret and ongoing “Islamization of the Western world” (e.g., Hafez, 2013) shows resemblance to anti-Semitic narratives (e.g., Zia-Ebrahimi, 2018) and is propelled by the idea of a demographic threat (e.g., Schiffer & Wagner, 2011). Some believers claim that there is a well-orchestrated campaign (e.g., “EURABIA”) to impose Islamic practices on the West. Such beliefs can have highly consequential outcomes. Indeed, they inspired the Norwegian terrorist, Anders Behring Breivik, to kill 77 people (Fekete, 2012; Uenal, 2016b).

Belief in conspiracies has been shown to be distinct from stereotypes or prejudices (Abalakina-Paap et al., 1999; Moscovici, 1987). They are also uniquely associated with negative psychological and behavioral outcomes (Bilewicz & Krzeminski, 2010), including political violence (Bartlett & Miller, 2010), and they show strong associations with anxiety, uncertainty, and a perceived lack of control (e.g., van Prooijen & Douglas, 2018; Whitson & Galinsky, 2008). Notably, in the context of anti-Semitism, Bilewicz and colleagues (2013) showed that conspiracy beliefs predicted discriminatory behavioral intentions, above and beyond prejudice alone. Based on this, we expect conspiracy beliefs to be statistically distinct from anti-Muslim prejudice and anti-Islam sentiment, to uniquely predict discriminatory attitudes and behaviors, and to be primarily associated with intergroup fear.

Emotional Underpinnings of Islamophobia: Who Is Afraid of Islam?

Another understudied issue regards Islamophobia’s emotional underpinning. Closely resembling concepts like Homophobia and Xenophobia, Islamophobia has been conceptualized from a standard social psychological research perspective, which puts fear and uncertainty about outgroups at the heart of the prejudiced mindset (e.g., Adorno et al., 1950; Duckitt, 2001; Jost et al., 2003; Wilson, 1973). This fear-primacy axiom is derived from research on the authoritarian character and the conservative mind more generally, which posits that political conservatism is embraced as an ideology because of its fear, anxiety, and uncertainty reducing qualities (Jost et al., 2003; Sibley & Duckitt, 2008). Indeed, the most frequently cited psychological definitions of Islamophobia hypothesize that fear is its main affective component (e.g., Conway, 1997; Lee et al., 2013). Yet, despite these claims, the role of fear in Islamophobia has not been systematically explored.

A socio-functional threat-based perspective on prejudice argues that different social groups can elicit different social threat perceptions, which in turn motivate specific attitudinal, emotional, and behavioral reactions to effectively mitigate perceived threats (e.g., Neuberg & Schaller, 2016; Park et al., 2007). For instance, Choma and colleagues (2012, 2016) have shown that intergroup disgust sensitivity is a robust predictor of Islamophobia. Given this, a socio-functional perspective may be better suited to understanding Islamophobia than focusing on fear alone. However, no previous studies have systematically analyzed multiple negative emotions (i.e., fear, disgust, and anger) concurrently and, thus, the emotional bases of Islamophobia and its subdimensions remain unclear.

Fear, Anger, and Disgust: The Potential Moderating Role of SDO and Ingroup Identification

A further understudied question is whether individual differences modulate the emotional reactions underlying Islamophobia (e.g., Choma et al., 2016; Matthews & Levin, 2012). Previous research has shown that group-based emotions can toggle between fear, anger, and/or disgust depending on a variety of other factors, including individual differences (for an overview see, Mackie & Smith, 2015) such as ingroup identification (e.g., Mackie et al., 2000) and SDO (Pratto et al., 1994). In many Western countries, Muslims and Islam are highly salient categories, frequently associated with negative connotations such as ideological/value conflicts, terrorism, as well as with realistic threats (e.g., to countries’ welfare systems; Bleich, 2011; Uenal, 2016a). Thus, individuals highly identified with their ingroup and/or high in SDO may be more likely to perceive Muslims and Islam as threatening, albeit accompanied with differing emotional reactions.

A significant body of research shows that negative outgroup affect is robustly predicted by SDO across various intergroup contexts (e.g., Sidanius et al., 2017). Individuals high in SDO are concerned with the maintenance of social hierarchies in general and are especially averse to low-status groups (e.g., Levin & Sidanius, 1999; Sibley & Duckitt, 2008). Importantly, some studies indicate that individuals high in SDO react with more anger and disgust, rather than fear, toward subordinate outgroups (e.g., Kossowska et al., 2008; Matthews & Levin, 2012) that are perceived to threaten the existing hierarchical order (Thomsen et al., 2008). More specifically, for individuals high in SDO, when Muslims and Islam are perceived as threatening, anger and disgust should be amplified to reinforce boundaries and preserve the hierarchical arrangement of society since these emotions are associated with behavioral tendencies which help facilitate actions to mitigate potential threats through aggressive means. Thus, anger and disgust are more reflective of a high SDO mindset toward threatening outgroups (e.g., Matthews & Levin, 2012; Sibley & Duckitt, 2008) while, for individuals low in SDO, given their more egalitarian worldview, the effects of existing anger and disgust should be buffered.

Another such variable that might explain individual differences in emotional reactions toward ethno-religious minorities is ingroup identification. Individuals who are highly identified with their ingroup tend to be more concerned about the well-being and distinctiveness of their own group and, in turn, more sensitive to ethnic/religious minority groups as potential threats (e.g., Riek et al., 2006) and more likely to react with negative intergroup affect (e.g., Mackie et al., 2000). The relationship between ingroup identification and specific emotional reactions seems less clear-cut than with SDO and has been shown to vary in conjunction with further variables such as perceived collective support (Dumont et al., 2003; Mackie et al., 2000). Thus, the specific affective response might vary independently of the strength of identification. Importantly, however, ingroup identification has been primarily linked to increased ingroup preference and love rather than outgroup derogation and aggression per se (Brewer, 1999; Levin & Sidanius, 1999). Thus, Muslims and Islam might trigger protective strategies to increase feelings of ingroup distinctiveness and facilitate ingroup solidarity. Against this background, it seems reasonable to assume that high ingroup identification might primarily moderate feelings of fear and disgust toward Islam and Muslims, such that higher fear and disgust might interact with higher ingroup identification as a means to signal ingroup distinctiveness and solidarity via outgroup derogation while, for individuals with low ingroup identification we expect that group-based fear and disgust might be less salient due to low group-based concern.

Overview of Studies

We had three main goals, which we tested across two studies, the first of which included five separate samples. First, to the best of our knowledge, there are no cross-national tests of the structure of Islamophobia. Thus, our first goal was to fill this research gap by cross-culturally validating our novel Islamophobia measure across five different contexts (i.e., United States, India, Germany, France, Poland). These five contexts were selected because they differ in characteristics such as culture, level of economic prosperity, relative size of Muslim population, history of conflict, and levels of immigration—all factors that may contribute to different understandings (i.e., measurement structures) of Islamophobia. Here, particularly analytic attention was given to testing the cross-cultural equivalence (or measurement invariance [MI]) of the scale or “whether or not, under different conditions of observing and studying phenomena, measurement operations yield measures of the same attribute” (Horn & McArdle, 1992, p. 117). Next, despite assumptions about fear in particular underlying Islamophobia, to our knowledge, this has also not been directly investigated. Thus, our second goal was to test the relationship of our measure with the intergroup emotions of fear, anger, and disgust. Given the cross-cultural nature of the present research, we were able to test whether the same emotions underpin Islamophobia across cultures.

In Study 2, we aimed to further investigate the relationship between intergroup emotions and Islamophobia by analyzing the role of two individual difference variables as potential moderators of this relationship in the United States. We also aimed to further gauge the scale’s psychometric qualities by testing its incremental validity. Here, we predicted that the Tripartite Islamophobia Scale (TIS) would predict negative intergroup outcomes such as dehumanization and ethnic persecution above and beyond well-established predictors. Furthermore, based on insights gained from research on anti-Semitism (e.g., Bilewicz & Krzeminski, 2010), we expected that the subcomponents of anti-Muslim prejudice and conspiracy belief would predict our criterion variables over and above anti-Islam sentiment.

For each study, the required sample sizes were calculated via power analyses (G*Power, Faul et al., 2009) for each statistical analysis individually to meet an 80% chance to observe a small to medium effect (f [2] = .04 to f [2] = .05). Detailed power analyses calculations for each statistical analysis are included in SOM (S1). All measures, conditions, quality checks, and data exclusions are reported in the respective method sections and in the SOM (see Supplemental Tables S1–S4).

Study 1

Methods

Participants

The present study was preregistered (see http://aspredicted.org/blind.php?x=6sk9gm). We aimed to collect 300 participants in each country (for power analyses, see S1 in SOM). In the United States (n = 295) and India (n = 293), the data were collected via Amazon MTurk. In Germany (n = 293), Poland (n = 297), and France (n = 288), the data were collected via the online participant recruitment service PROLIFIC. All samples were drawn between March and April 2019. The sample was 44.9% female and the average age was 36.4 (SD = 12.24). Detailed descriptive statistics for each country are provided in the SOM (Supplemental Tables S1–S5).

Instruments

TIS

Based on a preliminary study (Uenal, 2016c) and a review of existing academic and non-academic literature, 25 items were adapted and/or developed to measure the three proposed subcomponents of Islamophobia. This item selection was subsequently presented and discussed in various research labs in the United States, Germany, Austria, Poland, and France and with leading academic researchers in Islamophobia studies. This led to making further adjustments and selecting 15 items in total. The final items were translated into English, Polish, and French from German and back-translated by native speakers and/or professional translators (see Tables S14–S15 SOM for all translations). As discussed in the “Introduction,” the TIS was designed to capture three distinct aspects of beliefs and attitudes about Muslims and Islam (see Table S3 SOM). Cronbach’s alphas were satisfactory for each country and are listed in the SOM (see Table S4). Anti-Muslim prejudice (average α = .91), anti-Islam sentiment (average α = .91) Islamophobic conspiracy beliefs (average α = .95), and TIS aggregated scale (average α = .93), were each measured with five items (see Figure 1). These items were measured on a five-point scale ranging from 1 (definitively false) to 5 (definitively true). In addition to the scale, participants completed the following measures assessed on 5-point scales ranging from 1 (strongly disagree) to 5 (strongly agree), unless otherwise noted.

Figure 1.

The three-factor measurement model of the Tripartite Islamophobia Scale.

Note. The numbers are standardized factor loadings as estimated in a configural model with no constraints. The model χ2 of 719.837 indicates a lack of an absolute fit (p < .001), which is not uncommon for larger sample sizes (N = 1,466). However, all the other fit measures indicate that the model has a good model fit: χ2/df = 5.16; CFI = .98; TLI = .98; SRMR = .029; and RMSEA = .045 with 90% CI = [.041, .050]. CFI = comparative fit index; TLI = Tucker–Lewis index; SRMR = Standardized Root Mean Squared Error; RMSEA = root mean square error of approximation; CI = confidence interval.

Intergroup emotions

The intergroup emotions of fear, anger, and disgust were assessed with four items each framed toward Muslims and four corresponding items framed toward Islam (e.g., “I feel fearful when I think about Muslims [Islam],” “I feel repelled when I think about Muslims [Islam],” “I feel angry when I think about Muslims [Islam]”; Mackie et al., 2000). Due to length considerations, for the main analysis, we computed three variables, representing Fear of Muslims and Islam (average α = .97), Anger toward Islam and Muslims (average α = .97), and Disgust toward Islam and Muslims (average α = .97). Additional analyses analyzing each emotion relating either to Islam or Muslims individually, showed no overall differences and are provided in the SOM (for correlation analysis see Table S5; for regression analyses see Tables S6–S7).

Analysis of data

We first ran confirmatory factor analyses (CFA) in each country to test the proposed tripartite factor structure of Islamophobia. To this end, we tested if a three-factor model would fit the data better than a one-factor model, or theoretically justifiable two-factor models, in each analyzed country individually as well as in the overall sample. Next, we conducted multigroup MI analysis to test whether the proposed TIS showed cross-cultural equivalence. We then proceeded with the analysis of the emotional underpinnings of Islamophobia. We tested whether fear, anger, or disgust would be most predictive of our scale, by running multiple regression analyses and relative weights analysis (Tonidandel & LeBreton, 2015).

Results

Descriptive statistics

Table 1 shows the mean scores, standard deviations, and intercorrelations among the intergroup emotion variables and the Islamophobia subscales for the overall aggregate sample.

Table 1.

Means, Standard Deviations, and Intercorrelations Between All Variables Across Samples.

Variable	M	(SD)	1	2	3	4	5	6
1. Tripartite Islamophobia Scale	2.71	1.06	–
2. Anti-Muslim prejudice	2.60	1.12	.94***
3. Anti-Islam sentiment	3.12	1.13	.90***	.77***
4. Conspiracy beliefs	2.42	1.19	.93***	.83***	.72***
5. Fear Islam & Muslims	2.20	1.19	.76***	.72***	.66***	.72***
6. Anger Islam & Muslims	2.04	1.13	.81***	.79***	.70***	.74***	.79***
7. Disgust Islam & Muslims	1.86	1.10	.76***	.74***	.64***	.71***	.74***	.91***

*p < .05. **p < .01. ***p < .001.

Factorial structure and measurement equivalence

Confirmatory factor analysis

All CFAs were calculated using lavaan version 0.5–23 in R (Rosseel, 2012). A test of normality of the latent variables indicated a non-normal distribution. We, therefore, employed the Satorra–Bentler rescaling method for CFA estimation, as suggested by Rosseel (2012).

For the overall aggregate sample, the results revealed that the proposed three-factor solution yielded good fit to the data, χ2/df = 5.16; CFI = .98; TLI = .98; SRMR = .029; RMSEA = .045 with 90% CI = [.041, .050]; and AIC = 54,336.215, and indeed better than a one-factor, χ2/df = 18.85; CFI = .93; TLI = .91; SRMR = .48; RMSEA = .096 with 90% CI = [.092, .100]; and AIC = 55,509.398, and all possible two-factor solutions (see SOM Table S9). Next, the three-factor model also yielded better results for each country individually (see Tables S8–S9 in the SOM). Figure 1 depicts the results of a CFA with the confirmed three-factor structure for the overall aggregate sample.

MI analysis

Next, we tested whether our Islamophobia measure was cross-culturally equivalent by examining MI across the five samples. The primary objective was to ensure that the measurement models conducted under different conditions (geography, population, culture) yielded equivalent representations of the same construct (Jöreskog, 1971). The process of establishing MI includes six stages of analysis in total, of which only the first three steps are reported here due to space limitations (Steenkamp & Baumgartner, 1998). For further details on additional analysis (i.e., bi-factor model, second-order model), see caption of Table 2. By assessing whether factor loadings, intercepts, and residual variances are equivalent in a multigroup factor model (i.e., testing whether we measure the same underlying construct across the different contexts), we aimed to assure that comparisons that are made on our Islamophobia measure are valid across groups. If the first three levels of invariance—configural (same factor structure), metric (same factor loadings), and scalar (same intercept loadings)—across groups are met, comparison of the construct and item means across countries are justified. For comparison of effects, however, only the first two levels of invariance are required. Following Steenkamp and Baumgartner (1998), we adopted a “bottom-up” test strategy.

Table 2.

Multiple-Group Confirmatory Factor Analysis: Fit Measures and Differences of the Invariance Analysis.

Model	χ2(df)	Δχ2	CFI	ΔCFI	RMSEA	ΔRMSEA	SRMR	ΔSRMR
Configural invariance	909.198(410)	—	.980	—	0.051	—	0.036	—
Full metric invariance	989.972(458)	80.77***	.977	0.003	0.051	0.000	0.054	0.018
Full scalar invariance	1171.276(506)	144.20***	.970	0.010	0.056	0.005	0.058	0.022

Note. In addition, we also tested an alternative approach to the first-order factor measurement invariance analysis by also testing a second-order measurement invariance analysis and a bifactor model of our scale as comparisons. The results indicate that the fit of the first-order and second-order solutions fit the data equally good and better than the bifactor models. Results can be obtained by contacting the author. Robust SEM Model-fit (Satorra–Bentler correction; Mplus Variant). df = degrees of freedom; CFI = comparative fit index; delta CFI = difference in CFI from the previous model in the sequence; RMSEA = root mean square error of approximation, SRMR = standardized root mean square residual.

*p < .05. **p < .01. ***p < .001.

Chi-squared difference tests are generally recommended to test for MI by comparing nested models. However, the chi-squared difference test is influenced by sample size, almost always being significant in large samples (Chen, 2007). Therefore, we adopted a change in CFI between nested models of ≥0.010 in addition to a change in the RMSEA of ≥0.015 or a change in SRMR of ≥0.030 (for loading invariance) and ≥0.010 (for intercept invariance) as an appropriate criterion indicating a significant decrement in fit between models following Chen (2007). However, Chen (2007) suggested primarily using the change in CFI among the three indices for nested model comparisons as the other two are also affected by sample size.

Table 2 shows the results of a multigroup confirmatory factor analysis (MGCFA). The relative differences in CFI and RMSEA between the three models were below the required threshold of CFIdifference ≥ 0.010 and RMSEAdifference ≥ 0.015 (i.e., suggesting that they have equivalent fit to the data), indicating configural invariance, full metric invariance, and full scalar invariance. As a result, group comparison can be made on valid grounds.

Emotional underpinnings of Islamophobia

Having established cross-cultural equivalence at the metric level, the data further enabled us to test whether our Islamophobia measure is primarily associated with intergroup fear—as proposed by the most commonly referenced definitions (e.g., Conway, 1997; Lee et al., 2013)—or also by the intergroup emotions anger and/or disgust, as implicated by a socio-functional perspective (e.g., Neuberg & Schaller, 2016). To this end, we proceeded by testing whether fear, anger, and disgust are uniquely and significantly predictive of Islamophobia and to compare the relative strength of each intergroup emotion on our criterion variables.

Given the high intercorrelation between the intergroup emotions, we chose not to rely solely on standard multiple regression analysis due to known issues in over- and/or underestimating the relative effect sizes within settings of multicollinearity (e.g., Tonidandel & LeBreton, 2015). As an alternative we chose relative weights analysis (Johnson, 2000), which remedies multicollinearity issues by creating new sets of predictors, which are orthogonal to the original predictors and provide more accurate estimates of the relative strength of each predictor. Specifically, we tested and compared the relative weight of each intergroup emotion on the TIS as a whole as well as on each of its subcomponents following the approach outlined by Tonidandel and LeBreton (2015). Comparing these weights within the sample yields estimates about which emotion is relatively more important compared to others in explaining our criterion variables. We also included results from the traditional multiple regression analyses to display the direction of the relationships between intergroup emotions and our criterion variables.

Table 3 displays the results from the hierarchical regression and relative weights analyses. Due to length considerations, the results for the hierarchical regression analyses are presented in SOM (Tables S10–S11). Importance weights are interpreted as the proportion of each variance’s contribution to the criterion value (Tonidandel & LeBreton, 2015); therefore, higher relative weights are associated with a higher portion of a model’s R 2. The rescaled importance weights are the percentage of the importance weights’ contribution to a model’s R 2 value. Hence, a higher percentage is interpreted as a higher proportion of the total R 2 value. We included the beta values for comparison purposes and to assess discrepancies between them and importance weights.

Table 3.

Multiple Regression and Relative Importance of Intergroup Emotions on Islamophobia and Its Subcomponents in the Aggregated Sample (N = 1,466).

Country	Predictors	Tripartite Islamophobia Scale			Anti-Islam Sentiment			Anti-Muslim Prejudice			Conspiracy Beliefs
		ß	IW	RIW (%)	ß	IW	RIW (%)	ß	IW	RIW (%)	ß	IW	RIW (%)
5 Countries (N = 1,466)	Fear Islam & Muslims	0.32***	.23*	34.04	0.25***	.18*	34.05	0.26***	.21*	31.93	0.35***	.21*	36.03
	Anger Islam & Muslims	0.46***	.25*	36.10	0.50***	.20*	37.81	0.49***	.24*	37.34	0.30***	.20*	33.41
	Disgust Islam & Muslims	0.10**	.20*	29.86	-0.01	.14*	28.13	0.10**	.20*	30.73	0.18***	.18*	30.56
	R 2	.69			.52			.65			.60
	IW		.69			.52			.65			.60
	RIW			100			100			100			100
USA (n = 295)	Fear Islam & Muslims	0.30***	.24*	33.39	0.29***	.19*	33.93	0.21***	.20*	30.84	0.35***	.22*	35.37
	Anger Islam & Muslims	0.60***	.26*	36.90	0.63***	.22*	37.87	0.56***	.25*	37.53	0.48***	.22*	35.32
	Disgust Islam & Muslims	–0.03	.21*	29.71	–0.15	.16*	28.19	0.07	.21*	31.63	0.01	.18*	29.31
	R 2	.71			.58			.66			.62
	IW		.71			.58			.66			.62
	RIW			100			100			100			100
Germany (n = 293)	Fear Islam & Muslims	0.20***	.17*	25.52	0.15***	.13*	23.87	0.14**	.15*	22.79	0.25***	.13*	32.11
	Anger Islam & Muslims	0.37***	.25*	37.48	0.37***	.21*	38.67	0.41***	.25*	39.25	0.19***	.14*	33.26
	Disgust Islam & Muslims	0.34***	.25*	36.99	0.30***	.21*	37.46	0.32***	.24*	37.95	0.27**	.14*	34.62
	R 2	.67			.56			.65			.42
	IW		.67			.56			.65			.42
	RIW			100			100			100			100
Poland (n = 297)	Fear Islam & Muslims	0.32***	.20*	35.67	0.25***	.12*	36.13	0.30***	.19*	35.26	0.30***	.16*	35.34
	Anger Islam & Muslims	0.45***	.22*	38.10	0.40***	.13*	39.48	0.52***	.21*	40.46	0.29**	.16*	34.84
	Disgust Islam & Muslims	0.06	.15*	26.22	–0.01	.08*	24.39	–0.03	.13*	24.27	0.18*	.14*	29.82
	R 2	.57			.34			.53			.46
	IW		.57			.34			.53			.46
	RIW			100			100			100			100
France (n = 289)	Fear Islam & Muslims	0.31***	.24*	32.43	0.28***	.18*	34.95	0.24***	.21*	29.40	0.31***	.21*	33.04
	Anger Islam & Muslims	0.34***	.26*	34.51	0.43***	.19*	36.48	0.28***	.25*	34.45	0.21*	.21*	32.84
	Disgust Islam & Muslims	0.28***	.24*	33.06	0.04	.15*	28.56	0.38***	.26*	36.15	0.33***	.22*	34.12
	R 2	.74			.51			.72			.64
	IW		.74			.51			.72			.64
	RIW			100			100			100			100
India(n = 292)	Fear Islam & Muslims	0.20***	.21*	30.52	0.21*	.17*	31.37	0.08	.18	27.72	0.26***	.21*	32.36
	Anger Islam & Muslims	0.61***	.25*	36.93	0.64***	.20*	37.45	0.55***	.24	37.62	0.53***	.23*	35.75
	Disgust Islam & Muslims	0.04	.22*	32.55	–0.10	.17*	31.18	0.18	.22	34.66	0.04	.20*	31.89
	R 2	.69			.54			.64			.64
	IW		.69			.54			.64			.64
	RIW			100			100			100			100

Note. Importance weights are relative weights which sum to the R 2 of the full model. Rescaled importance weights are calculated by dividing the importance weight by the R 2 of the model. IW = Importance Weight; RIW = Rescaled Importance Weight.

*p < .05. **p < .01. ***p < .001.

In line with our expectations, all three emotions showed unique and significant contributions in explaining our criterion variables (see Table 3 and Tables S10–S11 in the SOM). More specifically, intergroup anger was predictive of Islamophobia, above and beyond fear, in the overall sample as well as in the individual country analyses in terms of explained variance. Intergroup disgust was predictive of Islamophobia in the overall sample and for Germany and France in the country-level analyses. For the United States, Poland, and India, disgust was not predictive of Islamophobia over and above fear and anger. Regarding the subcomponents, again, anger was predictive of each subcomponent of our scale across all five countries. Disgust was predictive of some subcomponents in Germany, India, France but not in the United States or Poland (see Table 3).

Taken at face value, the relative weight analyses indicate that anger showed slightly larger and more consistent relative importance weights and explained larger proportions of variance for the TIS as a whole (IW = 0.25, p < .05; RIW = 36%), for anti-Islam sentiment (IW = 0.20, p < .05; RIW = 37%), and anti-Muslim prejudice (IW = 0.24 p < .05; RIW = 37%), compared to fear and disgust. On the other hand, fear showed larger relative importance weight and explained larger proportions of variance in regard to conspiracy beliefs (IW = 0.21, p < .05; RIW = 36%).

Next, we tested whether the relative importance weights between the three emotions were significantly different from each other (see Table 4).

Table 4.

Comparing Relative Importance Weights of Intergroup Emotions in Predicting Islamophobia and Its Subcomponents Across All Samples.

Country	Reference Predictor Emotion (Intergroup Fear)	Tripartite Islamophobia Scale		Anti-Islam Sentiment		Anti-Muslim Prejudice		Conspiracy Beliefs
		LLCI	ULCI	LLCI	ULCI	LLCI	ULCI	LLCI	ULCI
5 Countries (N = 1,466)	Anger Islam & Muslims	−0.014	0.041	−0.007	0.045	0.006	0.063	−0.046	0.013
	Disgust Islam & Muslims	−0.055	−0.002	−0.055	−0.007	−0.036	0.020	−0.063	−0.003
USA (n = 295)	Anger Islam & Muslims	−0.039	0.079	−0.039	0.081	−0.006	0.096	−0.059	0.058
	Disgust Islam & Muslims	−0.083	0.011	−0.083	0.012	−0.043	0.053	−0.093	0.017
Germany (n = 293)	Anger Islam & Muslims	0.022	0.140	0.032	0.133	0.048	0.171	−0.064	0.072
	Disgust Islam & Muslims	0.014	0.146	0.015	0.137	0.034	0.171	−0.067	0.083
Poland (n = 297)	Anger Islam & Muslims	−0.064	0.092	−0.058	0.085	−0.053	0.103	−0.080	0.075
	Disgust Islam & Muslims	−0.124	0.019	−0.104	0.023	−0.130	0.013	−0.103	0.049
France (n = 289)	Anger Islam & Muslims	−0.032	0.088	−0.046	0.068	−0.024	0.103	−0.062	0.065
	Disgust Islam & Muslims	−0.053	0.082	−0.087	0.023	−0.011	0.126	−0.063	0.081
India (n = 292)	Anger Islam & Muslims	−0.008	0.097	−0.024	0.084	0.016	0.116	−0.034	0.071
	Disgust Islam & Muslims	−0.032	0.061	−0.049	0.045	0.001	0.093	−0.051	0.042

Note. The reference intergroup emotion fear is compared to anger and disgust. If zero is not included in the confidence intervals, weights are significantly different from one another. Significant differences are displayed in bold. LLCI = Lower level confidence interval; ULCI = Upper level confidence interval.

Fear and anger did not show significant differences in their relative weights in explaining TIS, 95% CI = [−.014, .041], anti-Islam sentiment, 95% CI = [−.007, .045], or conspiracy beliefs, 95% CI = [−.046, .013]. However, for anti-Muslim prejudices, the 95% CI = [.006, .063] did not include zero, and thus the results indicate that anger (IW = 0.24, p < .05, RIW = 37%) had a relatively larger weight size compared to fear (IW = 0.21, p < .05, RIW = 31%). In the country-level analyses, anger and disgust did now show significant differences in the United States, Poland, or France. However, both anger and disgust were stronger predictors, compared to fear, for the TIS, anti-Islam sentiment, and anti-Muslim prejudice, in Germany. Moreover, both anger and disgust were also significantly stronger predictors of anti-Muslim prejudice in India. Table 5 shows the comparison of relative importance weights with disgust as reference predictor.

Table 5.

Comparing Relative Importance Weights of Intergroup Emotions in Predicting Islamophobia and Its Subcomponents Across All Samples.

Country	Reference Predictor Emotion (Intergroup Disgust)	Tripartite Islamophobia Scale		Anti-Islam Sentiment		Anti-Muslim Prejudice		Conspiracy Beliefs
		LLCI	ULCI	LLCI	ULCI	LLCI	ULCI	LLCI	ULCI
5 Countries (N = 1,466)	Fear Islam & Muslims	0.001	0.055	0.006	0.055	−0.020	0.036	0.003	0.062
	Anger Islam & Muslims	0.027	0.059	0.035	0.066	0.026	0.059	0.000	0.034
USA (n = 295)	Fear Islam & Muslims	−0.023	0.077	−0.011	0.085	−0.055	0.041	−0.014	0.093
	Anger Islam & Muslims	0.018	0.009	0.023	0.098	0.010	0.078	0.005	0.076
Germany (n = 293)	Fear Islam & Muslims	−0.145	−0.014	−0.135	−0.016	−0.171	−0.033	−0.085	0.064
	Anger Islam & Muslims	−0.047	0.055	−0.040	0.054	−0.041	0.058	−0.053	0.048
Poland (n = 297)	Fear Islam & Muslims	−0.021	0.124	−0.019	0.107	−0.014	0.130	−0.047	0.102
	Anger Islam & Muslims	0.029	0.108	0.014	0.091	0.048	0.128	−0.021	0.065
France (n = 289)	Fear Islam & Muslims	−0.081	0.054	−0.023	0.087	−0.124	0.010	−0.080	0.063
	Anger Islam & Muslims	−0.021	0.040	0.011	0.074	−0.054	0.023	−0.048	0.031
India (n = 292)	Fear Islam & Muslims	−0.061	0.033	−0.047	0.049	−0.093	−0.001	−0.041	0.053
	Anger Islam & Muslims	0.002	0.073	0.008	0.070	−0.011	0.062	−0.002	0.059

Note. The reference intergroup emotion disgust is compared to anger and fear. If zero is not included in the confidence intervals, weights are significantly different from one another. Significant differences are displayed in bold. LLCI = Lower level confidence interval; ULCI = Upper level confidence interval.

In the overall aggregated sample, both fear and anger showed significantly larger associations compared to disgust, except for anti-Muslim prejudice (see Table 5). Regarding the individual country-analyses, anger was a significantly stronger predictor, compared to disgust for the United States and Poland, while fear and disgust did not differ for these countries. Disgust was also a stronger predictor compared to fear in Germany (except for conspiracy beliefs). Disgust was also relatively stronger compared to fear in predicting anti-Muslim prejudice in India.

Study 1 Discussion

Study 1 lends support for our hypothesis that Islamophobia is best explained as a three-factor model, consisting of anti-Muslim prejudice, anti-Islam sentiment, and Islamophobic conspiracy beliefs. The TIS also showed scalar invariance in the overall sample. Hence, the TIS seems appropriate for cross-cultural analysis, including the comparison of mean values. As for the second aim of exploring emotional associations with the TIS, the results were partially in line with our expectations that anger and disgust would be predictive of Islamophobia, above and beyond fear. In the aggregated sample analyses, fear, anger, and disgust were all uniquely associated with Islamophobia as a whole as well as its subcomponents. However, anger and fear did not differ in their relative weight in explaining Islamophobia as a whole. Regarding the analyses of the individual subcomponents, anger was significantly more predictive than fear of the anti-Muslim prejudice dimensions. Moreover, disgust was less predictive than fear of all, except the anti-Muslim prejudice dimension. In terms of country-level analyses, we found a more heterogenous picture which we discuss in more detail in the general discussion.

Study 2

The aim of Study 2 was twofold. First, we investigated the moderating role of two potential individual differences (i.e., ingroup identification and social dominance orientation) on the relationship between intergroup emotions and Islamophobia (e.g., Kossowska et al., 2008; Matthews & Levin, 2012). Second, we investigated a number of attitudinal and behavioral outcomes expected to differentially relate to each subcomponent of the TIS. To this end, we tested whether our Islamophobia scale would predict negative intergroup outcome variables as criterion variables while controlling for further related variables. Moreover, based on previous research on anti-Semitism (e.g., Bilewicz & Krzeminski, 2010), we hypothesized that prejudices and conspiracy beliefs would be stronger predictors of intergroup attitudes and behaviors focused on the overt domination and subjugation of Muslims (i.e., ethnic persecution) than anti-Islamic sentiment.

Amazon MTurk was used to collect a U.S. sample (N = 213) between March and April of 2019. The sample consisted of 46.9% women and the average age was 31.70 (SD = 1.17). Moreover, 14.6% had completed high school or less, 22.5% had completed some college, 37.6% had completed a bachelor’s degree, and 8.5% had partially completed or completed a graduate or professional degree. Detailed descriptive statistics for the sample are provided in the SOM (see Table S2).

The TIS developed in Study 1 was administered (anti-Islam sentiment, α = .91; anti-Muslim prejudice, α = .92; anti-Muslim conspiracy beliefs, α = .97; TIS aggregated scale, α = .96) together with the same intergroup emotions scales (fear of Islam and Muslims, α = .97; anger toward Islam and Muslims, α = .97; disgust toward Islam and Muslims, α = .97). For the TIS, confirmatory factor analysis replicated measurement structure from Study 1 (see Table S13 in SOM).

Individual difference measures

We administered two additional measures, both scored on a five-point scale ranging from 1 (strongly disagree) to 5 (strongly agree).

SDO

SDO was assessed through use of the 16-item (e.g., “Some groups of people are simply inferior to other groups”; α = .96) SDO-7 scale (Ho et al., 2015).

Ingroup identification

Ingroup identification was measured using five items (e.g.; “I feel strongly connected to other Americans”; α = .96) adapted from the collective self-esteem scale (Luhtanen & Crocker, 1992).

Predictive validation measures

To assess the incremental validity of our scale, we administered three additional measures. All predictive measures were scored on a five-point scale ranging from 1 (strongly disagree) to 5 (strongly agree) unless stated otherwise.

Ethnic persecution

Ethnic persecution was measured by an adapted version (Thomsen et al., 2008) of Altemeyer’s (1996) posse scale. Six items assess one’s willingness to participate in the politically sanctioned persecution of Muslims (e.g., “I would participate in attacks on Muslim headquarters if supervised by the proper authorities”; α = .93).

Anti-Muslim policy support

Support for anti-Muslim policies was assessed by nine items from Kteily et al. (2016; e.g., “We need to stop accepting Muslim refugees into this country, period”; α = .97).

Dehumanization

Dehumanization of outgroup members (i.e., Muslims) was assessed with the Ascent of Man scale (Kteily et al., 2015) in which participants rated the average “evolvedness” of Muslims from 0 to 100 (full humanity). We used a reverse score for analysis, with higher values indicating more dehumanization.

First, we tested whether SDO and/or ingroup identification moderate the relationship between the intergroup emotions and Islamophobia. Next, to test for the predictive validity of the TIS, we used hierarchical regression analysis by stepwise regressing three criterion variables (e.g., ethnic persecutions) on the three subcomponents of TIS and our individual differences measures.

Table 6 shows the mean scores, standard deviations, and intercorrelations among the criterion variables and the Islamophobia subscales. All the measures showed significant intercorrelations in the expected direction. Additional results are provided in the SOM (Table S12), investigating each intergroup emotion separately for Islam and Muslims.

Table 6.

Means, Standard Deviations, Scales, and Intercorrelations Between Variables (N = 213, United States: Sample 6).

Variable	M	(SD)	1	2	3	4	5	6	7	8
1. Social dominance orientation	2.01	0.97	—
2. Ingroup identification	3.73	1.07	.26***
3. Fear of Islam & Muslims	2.17	1.29	.44***	.32***
4. Anger Islam & Muslims	2.12	1.23	.50***	.27***	.83***
5. Disgust Islam & Muslims	2.06	1.26	.51***	.25***	.80***	.95***
6. Anti-Islam Sentiment	2.50	0.96	.36***	.34***	.64***	.77***	.73***
7. Anti-Muslim Prejudice	2.08	0.97	.52***	.37***	.73***	.79***	.78***	.79***
8. Conspiracy Beliefs	2.38	1.36	.47***	.34***	.71***	.80***	.78***	.81***	.87***
9. Tripartite Islamophobia Scale	2.35	1.04	.48***	.37***	.75***	.83***	.80***	.92***	.94***	.94***

*p < .05. **p < .01. ***p < .001.

Moderation analysis

To test whether ingroup identification and/or SDO would moderate the relationship between the three intergroup emotions and Islamophobia we ran a hierarchical regression analysis. In the first step, we entered both individual difference variables ingroup identification and SDO along with the intergroup emotions. In the second step, we additionally entered the interaction terms between the moderators (centered) and the intergroup emotions (centered). Table 7 shows a summary of the hierarchical regression analysis for Islamophobia.

Table 7.

Regression Analyses Predicting Islamophobia.

Model	Tripartite Islamophobia Scale
	B	SE(HC4)	T	p
Step 1
Ingroup identification	0.15	.05	2.72	.007
Social dominance orientation	0.08	.07	1.22	.225
Intergroup fear	0.11	.08	1.44	.152
Intergroup anger	0.51	.15	3.32	.001
Intergroup disgust	0.21	.14	1.45	.148
Step 2
ID	0.14	.06	2.41	.017
SDO	0.12	.07	1.76	.079
Intergroup fear	0.08	.08	0.97	.334
Intergroup anger	0.41	.16	2.55	.011
Intergroup disgust	0.34	.15	2.18	.030
ID × Fear	0.26	.09	2.88	.004
ID × Anger	−0.16	.14	−1.15	.250
ID × Disgust	−0.08	.13	−0.58	.565
SDO × Fear	−0.19	.07	2.61	.009
SDO × Anger	0.39	.19	2.00	.046
SDO × Disgust	−0.24	.17	−1.39	.166

Note. Continuous variables (ingroup identification, social dominance orientation, and intergroup emotions) are centered. Significant effects are displayed in bold. ID = Ingroup identification; SDO = social dominance orientation.

In the first step of the analysis, F(5, 207) = 81.1, p < .001, R 2 = .66, only ingroup identification and intergroup anger showed significant effects on Islamophobia. Entering the interaction terms in the second step of the analysis, F(11, 201) = 39.11, p < .001, R 2 = .68, increased the explained variance slightly by 2%. In the second step, intergroup disgust also showed a significant positive effect.[1] Regarding the moderation analyses, in line with our expectations, ingroup identification and fear showed a significant interaction, such that higher fear was associated with increased Islamophobia among those with higher ingroup identification (see Figure 2).

Figure 2.

Simple slopes for interactions between intergroup emotions and ingroup identification and SDO. Note. Ribbons present 95% confidence intervals. SDO = social dominance orientation.

Moreover, in line with our hypothesis, anger and SDO showed a significant interaction such that higher anger was associated with increased Islamophobia for individuals high in SDO. Unexpectedly, SDO also showed a significant interaction with fear, such that higher fear was associated with significantly higher Islamophobia for low SDO individuals. No interaction between disgust and SDO or ingroup identification was indicated.

Predictive validity

Finally, we regressed three negative intergroup outcome variables (dehumanization, support for anti-Muslim policies, and ethnic persecutions) on demographic indicators, ingroup identification, SDO, and our TIS subscales in hierarchical multiple regression analyses. This enabled us to test the incremental validity of our scale by controlling for predictors expected to be related to the criterion variables. In the first step, we entered demographic variables (age, gender, education, and political orientation), ingroup identification and SDO. In the second step, we entered all subcomponents of our Islamophobia scale (anti-Muslim prejudice, anti-Islam sentiment, and conspiracy beliefs) into the model (see Table 8).

Table 8.

Regression Analysis With the Three Subcomponents of the Tripartite Islamophobia Scale and Criterion Variables (N = 213, Sample 6).

Islamophobia criteria	Dehumanization			Anti-Muslim policy				Ethnic persecution
	ß	SE(HC4)	p	ß		SE(HC4)	p	ß		SE(HC4)	p
Step 1
Age	.19	0.06	.752	−.18		0.03	.581	−.05		0.04	.198
Gender	.13	0.14	.348	.15		0.08	.045	−.18		0.09	.048
Education	.00	0.05	.972	−.04		0.03	.187	−.04		0.03	.231
Political orientation	.05	0.06	.412	−.00		0.03	.957	−.03		0.04	.519
Ingroup identification	−.07	0.07	.338	.04		0.04	.300	.00		0.04	.962
SDO	.18	0.09	.049	−.18		0.03	<.001	−.05		0.04	<.001
Step 2
Anti-Islam Sentiment	.03	0.13	.795	−.13		0.07	.075	−.17		0.08	.037
Anti-Muslim Prejudice	.12	0.16	.452	.73		0.09	<.001	.52		0.10	<.001
Conspiracy Beliefs	.39	0.12	<.001	.37		0.06	<.001	.26		0.07	<.001
Adjusted R 2	.36				.83				.67
Change in adjusted R 2	.17				.37				.23
Standard error of estimate	.97				.53				.62
Significant F change	p < .001				p < .001				p < .001

Note. Values reflect β coefficients for the full model. Age: 1 (lower) - 5 (higher); Gender: 1 (Male) – 2 (Female); Education: 1 (low) - 12 (high); Political Orientation: 1 (left) - 5 (right). All constructs were entered stepwise in a two-step hierarchical regression model. Significant effects are displayed in bold. SDO = Social dominance orientation.

Consistent with our expectation, anti-Muslim prejudice and the conspiracy belief subcomponents were significant and unique predictors, controlling for all the remaining measures, and explained an additional 17% to 37% of the total variance, over and above the well-established measures in the model (see Table 8).[2] In line with our expectation, anti-Muslim prejudice and conspiracy beliefs outperformed anti-Islam sentiment in predictive power. Conspiracy beliefs were, moreover, significantly associated with dehumanization.

Study 2 Discussion

Study 2 tested two individual difference variables explaining distinct emotional reactions to Islam and Muslims. Furthermore, we analyzed the incremental validity of the TIS. As hypothesized, especially for individuals high in SDO, anger was associated with higher Islamophobia. This is in line with previous research indicating a close relationship between SDO and intergroup anger (e.g., Matthews & Levin, 2012). Moreover, as expected, social identification moderated the effects of fear on Islamophobia. Unexpectedly, the results indicated that SDO also interacted with fear, such that higher fear and low SDO interacted in predicting increased Islamophobia. However, no interaction with disgust was shown. This suggests that other potential moderating factors might be at play that need further exploration to explain the disgust-Islamophobia nexus. We will discuss these finding in more detail in the general discussion.

Regarding the incremental validity of our scale, we hypothesized and found support that, compared to anti-Islamic sentiment, both anti-Muslim prejudice and conspiracy beliefs had larger effects on behavioral inclinations that promote the active and forceful oppression of Muslims and Islamic organizations (e.g., support for anti-Muslim policies). This prediction was partly based on research on anti-Semitism, which showed that above and beyond anti-Semitic prejudices, anti-Semitic conspiracy beliefs were most significantly predictive of negative intergroup outcome measures (Bilewicz et al., 2013).

General Discussion

Since the introduction of the term about 30 years ago, several conceptualizations and models of Islamophobia have been proposed (Bleich, 2011; Klug, 2012). Yet, many theoretical and methodological issues have persisted including a lack of knowledge about Islamophobia’s (a) components, (b) cross-cultural validity, (c) emotional underpinnings, (d) relation to individual differences, and (e) societal consequences. In this paper, we aimed to address each of these issues. Across several cultures, we reliably distinguished between three different types of attitudes and beliefs about Islam and Muslims, using the TIS. Second, across cultural contexts, we systematically showed that both anger and—to a lesser degree—disgust predict Islamophobia in addition to fear. Third, we demonstrate that individual differences in the emotional reactions of anger and fear can be explained as a function of SDO and ingroup identification. Finally, we analyze the scale’s predictive validity and demonstrate its relevance in explaining important societal outcomes (e.g., blatant dehumanization, anti-Muslim policy support).

A New and Cross-Culturally Valid Islamophobia Measure

One of the more persistent issues regarding research on Islamophobia was the lack of a cross-culturally validated instrument to measure the phenomena across time, groups, and geographical regions. Therefore, we designed and tested our Islamophobia measure in five different cultural contexts. The results show that our TIS instrument indicates full scalar invariance. This suggests that the scale assesses the same underlying constructs in each setting and, hence, may be used to make valid comparisons, including among different mean levels, of Islamophobia across cultures.

Moreover, we contribute to current discussions of the comparability of Islamophobia with anti-Semitism and other prejudices, a topic which has triggered heated debates in the academic and public discourse, including questioning whether or not these phenomena are comparable to each other (e.g., Küntzl, 2008; Uenal, 2016b). Our results indicate that—comparable to modern anti-Semitism—Islamophobia manifest itself in different ways: negative sentiments against the Islamic religion, prejudice against Muslim individuals, and conspiracy beliefs about Muslims as collective agents with malicious and deceptive intentions. Our measure, thus, offers a conceptually sound way to compare Islamophobia with other forms of prejudices and conspiracy beliefs.

Conspiracy Beliefs as Component of Islamophobia

Our research advances theoretical discussions on Islamophobia by adding to a neglected, yet crucial, aspect of the concept: conspiracy beliefs. The results of this paper indicate that Islamophobic conspiracy beliefs were a statistically unique factor in five different cultural settings. We also found that the subcomponents disassociated in terms of the downstream intergroup outcomes they best predicted (i.e., ethnic persecution, dehumanization, anti-Muslim policy support). Similar to findings on anti-Semitism (Bilewicz et al., 2013; Bilewicz & Krzeminski, 2010), on which we partly based our hypothesis, we found that beyond anti-Islam sentiment alone, anti-Muslim prejudice and especially adherence to conspiracy beliefs were most predictive of these outcomes. Both of these subcomponents explained our three criterion variables over and above anti-Islamic sentiment. Crucially, our scale also showed good incremental validity and predicted the criterion variables, even when controlling for demographic variables, ingroup identification, and social dominance orientation. Moreover, we found that the conspiracy beliefs sub-component showed robust association with fear across the samples and, compared to disgust, slightly stronger associations with fear in the overall sample analyses. This finding aligns with previous experimental research showing that belief in conspiracy theories is associated with anxiety, uncertainty, and a perceived lack of control (e.g., van Prooijen & Douglas, 2018; Whitson & Galinsky, 2008).

Emotional Underpinnings of Islamophobia Across Cultures

Taking a socio-functional and threat-based approach to prejudice (e.g., Neuberg & Schaller, 2016), we hypothesized that anger and disgust would be predictive of Islamophobia, above and beyond fear. Indeed, we found that, controlling for intergroup fear, anger and disgust were unique and significant predictors of Islamophobia in the aggregated sample, explaining Islamophobia taken as a whole as well as its subcomponents, albeit to varying degrees. Nevertheless, applying a more stringent test to compare the relative strength of each emotion showed that anger and fear did not significantly differ in their relative predictive strength for the aggregated Islamophobia scale. However, anger was a significantly stronger predictor for the anti-Muslim prejudice component compared to both fear and disgust. Also, both anger and fear were significantly stronger predictors, compared to disgust, of the aggregated scale as well as two of its subcomponents. Thus, in addition to fear, considering both anger and disgust as integral parts of Islamophobia seems warranted.

In the country level analyses, the results were more heterogeneous. To start with, fear did not differ in relative strength from anger and disgust for the United States, Poland, and France. However, for Germany, we found that both anger and disgust showed significantly stronger associations with Islamophobia compared to fear. Anger, on the other hand, was significantly stronger compared to disgust for the United States, Poland and, partially, also for both France (anti-Islam sentiment) and India (TIS and anti-Islam sentiment).

Overall, this paper advances research on the emotional underpinning of Islamophobia by providing cross-cultural evidence for the relevance of the anger- and disgust-bases of Islamophobia. While our results indicate that fear does play a significant role in explaining Islamophobia, as suggested by the most prominent definitions and conceptualizations of current Islamophobia, our results ascribe anger an equally large role for explaining Islamophobia. Disgust, though significantly smaller in effect size, also played a significant role in predicting Islamophobia in the aggregated sample, and thus warrants more attention. Our findings on the country-level of analysis paint a more heterogeneous picture and point to context-dependent emotional underpinnings of Islamophobia. While it exceeds the aim and scope of this paper to explain the country-level variations that were not based on representative samples, they nevertheless raise intriguing questions as to how to explain such cross-cultural differences. For instance, in Germany, France, and also partially in India where anger and disgust were significant predictors of Islamophobia and most of its individual subcomponents, there are much larger Muslim populations (6.1% in Germany, 8.8% in France, and 15% in India; PEW, 2017b) compared to the United States and Poland (less than 1% of the overall population; PEW, 2017b). Given this, greater anger and disgust reactions in these samples might be a function of perceptions of increased socio-economic competition as well as perceptions of decreased collective support for the non-Muslim ingroup relative to population strength. Another interesting pattern which emerged is that anti-Muslim prejudice was more strongly associated with anger compared to fear, and in some countries, also stronger compared to disgust. Future research should further investigate these patterns and contextual differences.

Fear and Anger: Moderation by SDO and Ingroup Identification

This article also advances previous research by examining how differential emotional underpinnings of Islamophobia seem to partially serve as a function of psychological mechanisms (i.e., SDO and ingroup identification). Partially in line with our hypotheses, we found that individuals who scored higher on SDO showed more Islamophobia in conjunction with anger, but not disgust or fear. This is in accordance with previous research analyzing the associations of SDO and emotions in predicting negative intergroup outcomes, which points to anger as the primary affective component of an SDO mindset (e.g., Matthews & Levin, 2012). SDO has been consistently shown to be a robust predictor of negative outgroup attitudes and behavior signaling aggression toward low-status groups (Sidanius et al., 2017). An interesting, yet unexpected, finding of our study is the significant interaction between fear and SDO negatively predicting Islamophobia. More specifically, our results show that the effect of fear on Islamophobia seems to be buffered for high SDO individuals, while for low SDO individuals, fear leads to more Islamophobia. Low SDO has been previously associated with less negative intergroup bias. Individuals low in SDO are construed as more egalitarian and less dominant and as such, less prejudiced. However, our results indicate, if intergroup fear is present, low SDO can be associated with higher prejudice.

In contrast to SDO, we hypothesized that fear and disgust, but not anger, would interact with higher ingroup identification in predicting Islamophobia. Indeed, for individuals highly identified with their ingroup, we found that fear amplifies Islamophobic attitudes and beliefs. We interpret this finding in light of research which shows that ingroup identification is more indicative of ingroup love rather than outgroup derogation per se (Brewer, 1999; Levin & Sidanius, 1999). Individuals highly attached to their ingroup might adopt derogatory outgroup attitudes as a means of signaling their ingroup preference and thereby reasserting their distinctiveness and facilitating ingroup solidarity. Nevertheless, previous research has also shown that the association between ingroup identification and negative outgroup affect varies in tandem with other factors, such as perceived collective support (e.g., Mackie & Smith, 2015). Future studies might consider assessing both ingroup identification and perceived collective support as potential moderators to further explore the fear-Islamophobia nexus. Finally, we could not find an interaction with either of our moderators and disgust. Previous research has shown that right-wing authoritarianism (Altemeyer, 1996), an individual difference construct closely related, yet distinct from SDO, might be a better suited moderator between disgust and prejudice (e.g., Sibley & Duckitt, 2008).

Overall, our findings add to previous efforts of assessing ideological attitudes and social identification together with intergroup emotions in predicting specific intergroup outcomes (e.g., Kossowska et al., 2008; Matthews & Levin, 2012). In addition to corroborating previous results, our findings contribute to research by showing that low SDO can also be associated with higher prejudice, given the presence of intergroup fear. This is novel in light of previous research which has consistently found low SDO to be associated with less prejudice across different intergroup contexts (Sidanius et al., 2017). Our results indicate that an egalitarian and non-dominant worldview, as indicated by low levels of SDO, might not be sufficient to buffer the effects of fear on prejudice, at least not in the case of Islamophobia. Future studies could further explore this relationship.

Limits on Generality

While we believe that the cross-cultural approach taken in this study is a strength, the results should also be interpreted in light of some limitations. For instance, some socio-demographics varied between the countries (see Table S2 SOM). Moreover, although MTurk samples (and, presumably, PROLIFIC samples as well), are more representative than traditional social psychological samples (Paolacci & Chandler, 2014), these samples are not fully representative. In addition, concepts like Islamophobia are not static entities and develop over time. Thus, future research might reevaluate item selections depending on time, geography, and sociopolitical contexts, and potentially add further items to the scale. Indeed, the cross-cultural approach taken in the present research does not claim that the culturally most relevant or even most valid aspects of Islamophobia were assessed in each study. Although the item selection was based on extensive preliminary research and literature and was discussed with researchers from most of the countries in which data were collected, future research may profitably validate the content of the scales by, for instance, using qualitative methods as well. Besides others, this may allow a validation of the scale’s content against people’s evaluations, attitudes, and feelings regarding Muslims and their religion from a bottom-up perspective. Furthermore, future studies could consider behavioral validation methods, in addition to self-report validation approaches, to further gauge the psychometric properties of our scale.

Moreover, the three subcomponents of the TIS were strongly correlated. On the one hand, this may indicate the intertwined nature of Islamophobic attitudes and sentiments. On the other hand, future refinements of this scale may test whether rephrasing some of the items possibly reduces linguistic overlap between them, which may have led to higher inter-item correlations. Nevertheless, despite these findings, it is important to note that our proposed three-factor solution was confirmed across five culturally different countries supporting the tripartite nature of Islamophobia. Furthermore, criterion validation analyses further corroborated our tripartite view by showing theoretically justifiable, and statistically significant disassociation, of the three subcomponents in regard to their respective emotional antecedents as well as the intergroup outcomes they predict.

Practical Implications and Future Directions

An interesting area for future research highlighted by the present work concerns Islamophobic conspiracy beliefs. Our results indicate that the phenomenon is common across the analyzed cultural contexts and statistically distinct from the other two subcomponents. Yet, it remains an under-investigated phenomenon. Further research on the emotional bases of Islamophobia (and other prejudices) in different regional contexts could advance knowledge on the proximal, ultimate, and societal factors that may explain cross-country variations. Future studies could also investigate further individual and group differences as potential moderators of the relationship between intergroup emotions and Islamophobia. Specifically, analyzing the role of perceived collective support and threat perceptions in fear-modulation might prove fruitful.

We hope that by providing a more rigorous operationalization, as well as enhanced tools of assessment of Islamophobia and its substantive subcomponents, the present research can help facilitate further research concerned with the causes and consequences of Islamophobia.

Click here for additional data file.

Supplementary_Online_Files_Resubmission_PSPB-19-518 for The Nature of Islamophobia: A Test of a Tripartite View in Five Countries by Fatih Uenal, Robin Bergh, Jim Sidanius, Andreas Zick, Sasha Kimel and Jonas R. Kunst in Personality and Social Psychology Bulletin

Click here for additional data file.

Supplemental Material, Uenal_Online_Appendix for The Nature of Islamophobia: A Test of a Tripartite View in Five Countries by Fatih Uenal, Robin Bergh, Jim Sidanius, Andreas Zick, Sasha Kimel and Jonas R. Kunst in Personality and Social Psychology Bulletin