Party affiliation and support for Muslim newcomers: masked opposition in the Norwegian context Open Access

This work considers how anti-immigrant sentiment is selectively revealed, offering a new perspective on the role stigma plays in defining the political mainstream. By way of a survey experiment in Norway, we measure the gap between what people express and the beliefs they hold anonymously, assessing how supporters of distinct political parties situationally conform to expectations about the expression of intolerance. Using three frames, we contrast attitudes toward immigrants, refugees and more general types of new entrants to Norway. We conclude that the intolerance is more widespread than openly expressed sentiment would suggest due to the systematic masking of targeted opposition to Muslim newcomers. Covert expression, under conditions of absolute and permanent anonymity, demonstrate hardened sentiment among the mainstream right. In other words, a large shift in covert attitudes is best understood as a shift in deference to norms about the expression of targeted bias – in this case toward Muslim newcomers – rather than a large change in underlying, anonymously expressed attitudes.

Immigration plays an important role in a changing European socio-political landscape. In particular, the issue of immigration has undergirded political and social change, including populist right-wing political parties in a number of European contexts (Stockemer 2017). In this study, we focus on differences between overt and covert support for Muslim newcomers to Norway. Using a list experiment embedded in the Norwegian Citizen Panel (NCP), we manipulate the level of anonymity provided respondents. This approach allows us to consider differences in the extent to which support for Muslim newcomers is selectively expressed. Specifically, we are interested in the expectations that govern the overt expression anti-immigrant discourse, focusing on distinct types of Muslim newcomers.

Excellent work on attitudes toward immigrants – overall and Muslim newcomers in particular – has emerged in recent years (for summaries of the literature see Hainmueller and Hopkins 2014; Ceobanu and Escandell 2010). On the issue of Muslim newcomers, some have noted that Islam plays a unique role in shaping attitudes toward immigration in the Netherlands (Azrout and Wojcieszak 2017,¹; Erisen and Kentmen-Cin 2017; Savelkoul et al. 2011), Germany (Erisen and Kentmen-Cin 2017), the US (Creighton and Jamal 2015; Kalkan et al. 2009) and multi-country comparative work (Strabac et al. 2013; Strabac and Listhaug 2008). Characteristics of members of some contexts of reception, such as religiosity, has shown a significant and substantive link to anti-immigrant sentiment in Europe (Davidov et al. 2008; Sides and Citrin 2007). When considering characteristics of Muslim immigrants, findings vary. Some (e.g. Helbling 2014) find a link with general antipathy toward immigrants, but not a clear link with specific phenotypical symbols. Strabac et al. (2016), who consider the specific case of Norway, find that some expressions of religiosity among Muslim immigrants, namely the wearing of the veil, elicit targeted negative reactions that are distinct from general anti-immigrant sentiment. Recent work (Coenders et al. 2017) has considered distinct types of Muslim immigration (e.g. immigrants vs. refugees) and finds that the flow of asylum application, rather than general antipathy, is relevant.

To be clear, the approach used in this work is different than that intent on explaining general variation in attitudes toward Muslim immigration (i.e. the levels of opposition). Instead, the contribution and innovation of this work is a focus on how support/opposition to Muslim newcomers is selectively expressed. This insight is not wholly unrelated to observational approaches, but unique patterns can emerge. For example, work in the US has shown identical levels of opposition to Muslim and Christian immigrants is found only when respondents are able to mask their responses (Creighton and Jamal 2015). Rather than considering distinct religious groups, we reveal systematic differences in over-reporting of positive attitudes by type of Muslim newcomer and political affiliation of the respondent.

To these ends, Norway offers the opportunity to explore the overlap between political affiliation and attitudes toward Muslim newcomers as it sustains two large traditional/conventional parties on either side of the left-right divided – the Labour Party and the Conservative Party – and, moreover, a non-traditional, large and established party – the Progress Party. Importantly, the Progress Party is strongly focused on issue of immigration. This political landscape provides the key ingredients to explore distinct understandings of what is acceptable to express in terms of attitudes toward Muslim newcomers.

When comparing overt and covert expression of support for Muslim newcomers, our results reveal somewhat unexpected pattern with clear social and political consequences. Specifically, supporters of the centre-right Conservative Party exhibit substantive and significant masking of intolerance with relatively large gaps between what is overtly and covertly expressed, which reveals an interesting tendency: Conservative Party supporters are notably closer to Labour Party supporters on the issue of Muslim immigration in terms of their overt expression of support, regardless of the how the newcomers are framed. However, once anonymity is provided, a less tolerant position emerges placing Conservative Party supporters closer to the more restrictive stance of the Progress Party.

To better understand how context and political orientation shape the expression of attitudes toward Muslim newcomers in Norway, we consider the role of social desirability in stigmatising certain expressed attitudes and, moreover, how this might change under conditions of anonymity. Specifically, via the anticipation and avoidance of social stigma, group members (e.g. supporters of a political party) conform to situationally-specific role expectations based on the constellation of identities/roles/affiliations they embody. In a given interaction, participants make an effort to present a consistent self, pursuing a logic of situational conformity² (Blumer 1969; Goffman 1959; Kuhn 1964; Stryker 1980). However, this is not always possible and masking of identities anticipated to elicit greater stigma results. Succinctly put, ‘[r]ole conflict exists when there are contradictory expectations that attach to some position in a social relationship, [which] … may call for incompatible performances … Stryker (1980; 73).’ In other words, masking emerges as a strategy if (1) the potentially stigmatizing attribute is discreditable³ and (2) the associated stigma is able to be anticipated.

Empirical evidence has emerged of significant and substantive masking in a variety of social situations (Phillips and Clancey 1972; Presser and Stinson 1998; Arnold and Feldman 1981; Kuklinski et al. 1997a, b; Davis and Silver 2003; Kuran and McCaffery 2008; Heerwegh 2009). At the core of this work are findings that controversial attitudes, such as targeted opposition to specific immigrant groups, are often expressed only under conditions of absolute and permanent anonymity in a variety of contexts (Janus, 2010; Creighton and Jamal 2015; Creighton, Jamal and Malancu 2015; Knoll, 2013a, b; Sniderman and Hagendoorn 2007).

Of note, masking emerges even when the mode of data collection avoids direct interaction (e.g. phone, in-person interviews). For example, work in the US (Janus 2010; Creighton and Jamal 2015; Creighton et al. 2015) and the Netherlands (Creighton et al. 2019) finds significant masking using online surveys. D’Ancona (2014) assesses the link between the mode of collection and the expression of xenophobia, finding that less direct interactions reduce, but do not eliminate, social desirability bias. In summary, the mode of data collection is relevant and indirect approaches do result in a greater expression of controversial opinions such as intolerance toward immigrants. However, participation in a survey – whether online or not – results in some hesitancy to express some attitudes and approaches that provide anonymity in a permanent and absolute manner have demonstrable advantages if the goal is to capture covert sentiment.

We suggest that situational conformity, which is manifested in strategic masking of intolerance, offers a plausible theoretical frame by which intentional masking of antipathy toward newcomers can be anticipated. Specifically, we contend that the pursuit of a masking strategy, rooted in a sensitivity to conventional norms (i.e. what is and is not acceptable to overtly express) shaped by the anticipation of stigma, is an important feature of Norwegian politics. In practical terms, this should manifest itself in significant differences in stated attitudes by supporters of conventional parties (e.g. Conservative and Labour), depending on whether social stigma was present. In other words, we expect that supporters of the conventional parties will make a significantly greater effort to be consistent in their presentation of self, but a variation could emerge depending on the context/situation of a given interaction, particularly when anonymity is provided. In general, the logic of stigma predicts that supporters of unconventional/populist political parties being less concerns with complying with social norms about the expression of anti-immigrant sentiment.

Empirically, political orientation has been linked to antipathy/support for immigration in a variety of context (Sniderman et al. 2000; Betts 1988; Espenshade and Hempstead 1996). In general, research suggests that parties and their constituent supporters on the left and centre-left express less opposition to immigrants and immigration relative to their equivalents on the right and centre-right. A primary explanation is that that the underlying political ideology delineating the right and left, albeit consisting of substantial variation across contexts, does broadly reflect sincere differences in moral framing of the issue of immigration. On the left, progressive notions of inclusivity and a somewhat recent affinity with internationalism embraces and/or encourages followers who see immigration and immigrants as being encapsulated by broader tenants of inclusivity. In contrast, on the right, conservative notions of national and sociocultural continuity result in a scepticism of the compatibility of newcomers with national identity, national economies and sociocultural attributes (e.g. religion, ethnicity, race) (Isernia and Smets 2014; Sniderman et al. 2000).

Supporters of unconventional political movements, whether they be characterised as populist, radical, new right (Seippel and Strandbu 2016), populist radical right (PRR) (Muis and Immerszeel 2017; Golder 2016; Greven 2016; Mudde 2012) or populist radical right parties (PRRP) (Röth et al. 2018), are theoretically unmoved by situational norms governing role consistency on the right or the left. The term ‘populist’, as applied to these parties (Mudde 2007), reflects notably inconsistent positions in relation to other conventional left-right norms such as economic and social policy. That said, these same parties demonstrate notable consistency in terms of opposition to immigration (Van der Brug et al. 2000), offering one of the few unifying policy positions (van de Brug et al. 2005; Arzheimer 2009; Hainsworth 2008; Hutter and Kriersi 2013). Some suggest that the coupling of right-wing identification with strong anti-immigrant positioning defines alternatives to conventional political orientations (Van Spanje 2011). That said, variation does exist. Cases in point are contexts where anti-immigrant parties differ on support for redistributive welfare regimes (Finseraas 2012) and some Eastern European/post-Communist states (Van Spanje 2011).

As in other Western societies, political party affiliations in Norway correlate with anti-Muslim and anti-immigrant attitudes. At the moment of writing, there are nine political parties represented in Norwegian Parliament, with the three largest consisting of the Labour Party (‘Det norske arbeiderpartiet’; AP), the Conservative Party (‘Høyre’; H) and the Party of Progress (‘Fremskrittspartiet’; FrP).⁴ These three parties have consistently captured two-thirds of total vote in each election over the past 40 years (see Table 1). AP is a centre-left party that has dominated Norwegian politics for decades and remains the largest party despite being out of power since 2013. H is a conventional centre-right party that has traditionally been one of the largest parties in Norwegian politics and is currently the largest member in the governing coalition, determining the Prime Minister.

FrP is a relatively more recent entrant in Norwegian politics. The party emerged from an anti-tax party in 1973 with the rather long title ‘Anders Lange’s Party for a Strong Reduction in Taxes, Duties and Public Intervention’ – shortened to the Progress Party (FrP) in 1977. H formed a governmental coalition with FrP after the 2013 parliamentary elections, building upon earlier experience of working together in local councils (Jupskås 2013). With support from smaller, centre-right parties, the two parties have governed uninterrupted since 2013, repeating their electoral success in the 2017 elections (see Heinze 2018 for an overview). Frp is currently the third-largest party in the Norwegian parliament, receiving 15.2 percent of the votes in 2017 parliament elections (see Table 1).

Regarding attitudes toward immigrants and immigration, there have historically been clear differences between the AP, H and FrP – reflected in party platforms, public discourse and attitudes of their consituencies (Simonnes 2013). FrP is the most stridently anti-immigrant of the three, promoting its position on this issue as core to its identity (Kestila and Søderlund 2007). FrP has actively sought an anti-immigrant constituency for decades and, within a broader anti-immigrant posture, has consistently expressed targeted opposition toward Muslim immigrant population in particular (Strabac and Valenta 2012). The party has been strongly focused on immigration since the 1980s and was among the first to politicise the issue of immigration in Norway. (Kestila and Søderlund 2007). That said, the electoral success of FrP is not based solely on the issue of immigration as the party embraces a broader range of policies constituting a ‘ … rather erratic mixture of neo-liberalism, conservativism and populism’ (Hagelund 2003: 47). Nevertheless, the defining issue of the party is immigration. Despite being relatively more moderate, both H and AP recently advocated stricter immigration policies, due, at least in part, to the perceived electoral success of FrP on this issue (Simmones 2013; see van Spanje 2010 for broader European analysis of immigration and non-traditional party success).

We offer a number of testable hypotheses. Regarding the openly expressed opposition, we expect it will be significantly and substantively higher among supporters of FrP. In addition, we expect that supporters of centre-right H will express less support for Muslim newcomers than the supporters of centre-left AP. The difference between these two parties is expected to be smaller, compared to FrP. In other words, we expect voters of FrP to stand out as exhibiting particularly little support.

Both these expectations are largely based on findings from previous work.

When the threat of stigmatisation is minimised by the offer of absolute anonymity, significant change should emerge in expressed levels of support f only among members of conventional political parties – H and AP. We consider three categories of Muslim newcomers – Muslim refugees (MR), Muslim people (MP) or Muslim immigrants (MI). We expect masking to be basically non-existent among party supports of FrP, which is the clear marker that this party deviates from normative pressure and the situational conformity shaping the conventional Norwegian political parties such as H and AP. This expectation can be articulated in two, related hypotheses.

The item-count technique ICT, adapted for the substantive focus of this work (i.e. attitudes toward Muslim newcomers), works by manipulating the extent to which respondents are guaranteed absolute and permanent anonymity. First, a control group is presented a list of items unrelated to concerns about immigration about which they list the number with which they agree. Second, an independent treatment groups are presented with the same list, but with additional focal items, defining support for Muslim immigrants, refugees and people (i.e. newcomers). The difference between the mean response to the control list (first step) and the mean response to the treatment list (second step) offers a group-level measure of agreement with the focal item under conditions of absolute and permanent anonymity, termed covert sentiment. Third, the control sample is further sorted into randomised subsamples to which a direct question is presented that is directly comparable to one of the focal items. This step allows for the measurement of overtly expressed opposition, which is comparable to a standard public opinion poll.

[control list] Of the following three statements, HOW MANY of them do you AGREE with? We don’t want to know which statements, just HOW MANY.⁵

The treatment group is sorted further into three independent experimental subgroups. Each of these groups is presented with a similar list question as that received by the control group, but with an additional focal item added to the list. Each treatment subgroup, to offer sufficient statistical power, is roughly equivalent in size to the total control group. In total, three focal items are used – one per treatment subgroup – which constitute 4-items lists and are worded as follows:

[treatment list frame 1] Norway should allow people from Muslim countries to come and live here

[treatment list frame 2] Norway should allow immigrants from Muslim countries to come and live here

[treatment list frame 3] Norway should allow refugees from Muslim countries to come and live here

All control respondents are confronted with the 3-item list, but the control group is further divided into three independent experiment subgroups. Each of these subgroups is presented with a direct question that is directly comparable to one of the focal items placed in a treatment subgroup. This provides the overt measure of attitudes toward each of the three Muslim newcomer frames – immigrant, refugee and person. The wording of the overt questions is as follows:

[control direct frame 1] Should Norway allow people from Muslim countries to come and live here?

[control direct frame 2] Should Norway allow immigrants from Muslim countries to come and live here?⁶

[control direct frame 3] Should Norway allow refugees from Muslim countries to come and live here?⁷

In sum, the design used offers four distinct manipulations and related comparisons. The first is the difference between the treatment and control list, which reveals attitudes covertly expressed. A formal test of this difference is a one-sided test of means as the mean response to the 4-item treatment group, which includes the additional focal item, is expected to be equivalent or greater than the mean response to the 3-item control list. The second is the difference between each covert measure, which is calculable for Muslim newcomers characterised by three distinct frames – immigrant, refugee and person. There are two options to ascertain difference between each of the treatment frames. Either a two-sided test of proportions or a direct test of means between the mean response to the 4-item list. The third is the comparison between each overt measure of attitudes toward the same three frames. This is determined by a two-sided test of proportions. The fourth is the comparison between each overt frame and the equivalent covert frame (e.g. overt attitudes toward Muslim refugees vs. covert attitudes toward Muslim refugees). This is determined by a two-sided test of proportions.

The experiment described above was embedded in the Norwegian Citizen Panel (NCP)⁸, which is an ongoing representative panel survey of the Norwegian population jointly administered by Faculty of Social Sciences at the University of Bergen and Uni Research Rokkan Centre. The recruitment of panel members and the collection itself was completed in partnership with ideas2evidence⁹, a private data collection company, based on a sample drawn by Evry¹⁰, a private company that co-manages the National Registry of Norway with the Norwegian Tax Administration. The collection of 7th wave resulted in a total sample of 4,689 completed surveys, resulting in a cumulative response rate of 72 percent (Skjervheim and Høgestøl 2016). Of these 4,689 sampled individuals, a random subsample (n = 3,005) were included in the survey experiment used in this work with only respondents who identified with one of the three largest political parties – AP (n = 952), H (n = 692) or FrP (n = 281) – being retained in the analysis. The final analytic sample consists of 1,929 respondents and Figure 1 shows the process of randomly assigning these sampled respondents into the treatment and control groups. See Appendix 2 for a detailed overview of the NCP, analytic sample characteristics and mode of data collection.

The list experiment offers high levels of anonymity, which underpins its ability to mitigate the stigma associated with certain responses to survey questions. The maintenance of anonymity requires a number of assumptions. First, a response to a list question should not be zero. This is referred to as a floor effect and it would reveal a respondent’s preference for any item in the list (Blair and Imai 2012). Second, a response to a list question should not equal the total number of items. This is referred to as a ceiling effect and, like a floor effect, reveals the respondent’s preference for any given item – including the focal item (Blair and Imai 2012). Third, the difference in the number of items in the control list and the treatments list (control list + a focal item) should not affect the response pattern independent of a respondent’s response to the additional focal item. This last concern is a design effect (Blair and Imai 2012).¹¹

Tables A1, A2, A3 and A4 of Appendix 3 report the distribution of responses to the treatment and control list questions overall and by political party affiliation. The proportion and counts for responses, which range from zero to the total number of items (three in the case of the control and four in the case of the treatment), offer some indication of the extent to which there are respondents responding in a way that would violate the assumption that anonymity is provided by simply requiring a count response to a list rather than to individual items.

Two patterns emerge. First, there is little evidence of a ceiling effect in the sample overall (appendix Table A1) and for any political party (appendix Tables A2, A3 and A4). No more than seven respondents respond agree with four items regardless of the whether the overall sample or a specific political party are being considered. The pattern of ceiling effects varies little with 1–2 percent of respondents agreeing with all four items regardless of party affiliation. To be clear, this pattern is only suggestive as the possibility remains that a respondent could chose to mask their true response to avoid hitting the ceiling, which would show up in the data as a response less than four.

Second, between 7 and 21 percent of respondents to the treatment lists report a response of zero, indicating a non-trivial number of responses that reveal no agreement with any list item (i.e. floor effect). Moreover, within samples – for the total sample and political-party subsamples – the percentage responding zero to the treatment question are notably consistent, varying by no more than two percentage points.

The implications for the interpretation of the results for the H and FrP is twofold. First, both parties are overtly more restrictionist and, therefore, their supporters are less likely to agree with the additional focal item. Given the way in which the question is asked (i.e. agreement indicates greater support for Muslim newcomers), fewer zero responses due to residual social desirability bias not mitigated by the list question could result in an overestimate of the level of support. In other words, rather than report a zero, which reveals the response to any given item, respondents could suggest support that would not be the case if support could be withheld and anonymity maintained. From this perspective, the estimated lack of covert support for Muslim newcomers among those who identify with the FrP and H, described in detail below, should be interpreted as a conservative estimate.

Blair and Imai (2010, 2012) offer a way to assess the extent to which floor and celing effects could alter responses. The approach estimates the percentage of the sample/population that is likely to report a response other than that which would have emerged if anonymity were maintained – termed ‘liars’. Table A1 reports these estimates for each Muslim newcomer frame (i.e. people, immigrants and refugees) and for a combined sample of all respondents who received a treatment list question regardless of frame. The results suggest that the percentage remains below 1%, which suggests that here is little direct evidence that the floor effects generate bias in the reported.

Table 2 reports the n, mean and standard deviation of support for Muslim newcomers by frame (i.e. people, immigrants, people) and political party (i.e. FrP, H and AP). The estimated means/proportions or, in the case of covert support, the difference in means are formally tested. In addition, social desirability bias (SDB), which is the difference between the overt and covert estimates for each frame and political party, is reported and tested for significance.

Column d of Table 2 reports the overt support (rows e, g and i) for each frame for the entire sample, generalisable to the Norwegian population. A clear pattern emerges in that support is high when overtly expressed (Table 2; rows e, g and i), with more than half the population agreeing that Norway should allow Muslim newcomers, regardless of the frame, to come and reside. However, support overall in Norway is distinct from within-part support. AP supporters (Table 2; column b; rows e, g and i) and H supporters (Table 2; column c; rows e, g and i) are broadly similar to the overall population in terms of their overt attitudes toward Muslim newcomers, although AP supports consistently articulate slightly higher support than those identifying with H.

Supporters of FrP (column a) deviate from the general pattern observed for AP and H, with overt support being notably lower. No overt estimate of support for Muslim newcomer among FrP supporters significantly differs from zero (Table 2; column a; rows e, g and i), but this is to a large degree a result of small number of respondents in these cells Looking at the mean values of overt support, they are low, but not extremely so: The support varies from 10% for ‘Muslim immigrants’ to 26% for ‘Muslim refugees’.

In terms of the first set of hypotheses, we find support for both. Results show that, indeed, Muslim refugees, people and immigrants confront significantly less acceptance among FrP supporters relative to H and AP when sentiment is overtly expressed (H1a). In addition, Muslim refugees, people and immigrants confront a lower level of acceptance among supporters of H, compared with supporters of AP, when sentiment is expressed openly (H1b). In other words, in terms of overt support, we find it to be highest among AP supporter, followed closely by supporters of H. Those with an affinity for FrP are, as expected, a distant third.

When considering covert sentiment and SDB, the pattern deviates notably. Once permanent anonymity is provided by the ICT, covert support of H voters drops strongly (Table 2; column c; rows f, h and j). The levels of support are more than halved, and the decreases are statistically significant for all three wordings of the question. Regarding AP supporters, the results are somewhat more mixed. Also here, the general trend is decrease of support once the anonymity is provided, but the sizes of the decreases and their statistical significance are to some degree dependent on wording of the question. For the ‘Muslim people’ wording, the support drops by 20 percentage points, and the decrease is statistically significant. For ‘Muslim immigrants’ wording, the drop is about 16 percentage points, and with our sample sizes the decrease in support is not significant at the conventional 0.05 level (although it is significant at the 0.1 level). Finally, for the ‘Muslim refugees’ wording, the drop in support is about 10 percentage points and is not statistically significant. The results for supporters of FrP (column a), which need to be considered as suggestive due to the available sample size, reveal a complex pattern. For ‘Muslim people’ wording, the results are in line with our expectations – there is essentially no difference in support expressed openly and under anonymity condition. The ‘Muslim immigrants’ wording results in drop in support of about nine percentage points; a relatively modest, although non-negligible drop that is not statistically significant at our sample sizes. Finally, with regard to ‘Muslim refugees’ wording, the estimated drop in support is about 25 percentage points. That is definitely not a small decrease, although it is not statistically significant at the conventional 0.05 level due to low number of observations in cell (i) of the row (a).¹²

Thus, our hypothesis H2a (‘Muslim newcomers encounter significantly less tolerance among supporters of Labour (AP) and Conservative (H) parties when sentiment is expressed with the guarantee of absolute and permanent anonymity than when the same sentiment is overtly expressed’) receives mixed support. H supporters indeed show much less support in anonymity condition, while the results for AP supporters are mixed and the decrease in support in anonymity condition is generally less dramatic. Regarding the hypothesis H2b (‘Overt and covert expressions of tolerance for Muslim newcomers among supporters of the Progress Party (FrP) do not differ’), we cannot claim to have found any support for it, based on our dataset. Our results paint a pretty complex picture, where wording of the question seem to be of importance. This, combined with the low sample sizes in some of the table cells, prevent us from drawing any clear conclusions.

Table 3 reports the formal tests between AP, FrP and H supporters. Overall, H, FrP and AP supporters report significantly different levels of overt and covert support for Muslim Newcomers – framed as people (MP) and refugees (MI). In addition, the extent to which sentiment is masked (i.e. SDB) is notably different between distinct party affiliations. However, two exceptions are observed.

First, AP and H supports are notably similar in their overtly stated support for Muslim refugees (MR), but a large and significant gap emerges when anonymity is provided. In other words, what appears to be similarities between the two mainstream parties (AP and H) is limited to overtly expressed sentiment. Covert expression suggests a wider, covertly expressed difference between the two parties. Second, AP and FrP supporters do not mask much antipathy toward Muslim immigrants. This does not mean that they express similar levels of overt and covert support, which is not the case. Instead, the difference between the relatively positive overt and covert estimates for AP is similar to the difference between the relatively negative overt and covert estimates for FrP. In other words, AP supporters do not mask their support while FrP supporters do not mask their antipathy.

Of note, when comparing H and FrP supporters, the difference between support for Muslim newcomers across all frames narrows when sentiment is expressed covertly. An example is the decline from 0.474–0.176 in the case of support for Muslim immigrants (MI). This is not the case when results for AP supporters are compared to the FrP where a difference of 0.592 reduces to 0.515 for the same frame. In other words, H and FrP supporters appear distinct particularly when overt sentiment is assessed, but the provision of anonymity suggests the gap to be less substantive.

Our empirical results based on direct questions (i.e. overt) are very much in accordance with previous studies of attitudes toward new minorities and newcomers in Norway and, more broadly, Europe. Succinctly put, the levels of intolerance are notably higher among supporters of FrP than among supporters of centre-left (AP) and centre-right (H). As expected, some difference is also observed among conventional centre-right and centre-left party supporters with H expressing less support for newcomers than AP. That said, the two conventional parties are much closer to each other than to supporters of FrP in terms of overt attitudes.

These empirical results reinforce previous work that links overtly intolerant parts of population with support for populist-right parties and politicians. It is important to note that the sizable fraction of the electorate supporting overtly anti-immigrant parties in Norway (and Western Europe more broadly) do not comprise the majority of the population. Most of the electorate expressed relatively tolerant attitudes, which suggests that the growth of parties like the FrP based on the issue of immigration, in the absence of disruptive events (e.g. return to recession or dramatic shift in labour/refugee immigration), could be nearing its limits.

In contrast to the overt pattern, supporters of the H express significantly less support for Muslim newcomers when provided absolute and permanent anonymity. This pattern confirms the theoretical importance of situational conformity in that we interpret this pattern as emerging from the anticipation of the stigma. The proportion of the centre-right Conservative party’s supporters exhibiting support for Muslim newcomers, whether they be identified as immigrants, refugees or people, is more than halved when anonymity is provided; from about 60 to less than 30 percent. This effectively places H supporters, in terms of covert attitudes, closer to the supporters of the FrP than to the AP. Simply put, the attitudes of supporters of the centre-right party (i.e. H) might be less liberal than standard public opinion polls would suggest.

Regarding the supporters of the centre-left AP, our results are mixed and more dependent on wording of the question when it comes to decrease in support in conditions of anonymity. Particularly when asked about acceptance of ‘Muslim Refugees’, the AP-supporters tend to give similar answers, with difference between overt and covert condition being less than 10 percentage points and not statistically significant. Interestingly, quite the opposite is the case for FrP-supporters in ‘Muslim Refugees’ wording condition. For them the overt support increases, compared with ‘People’ and ‘Immigrants’ wordings, while the support under condition of anonymity drops to basically zero. This results in a fairly large difference between overt and covert condition (SDB) of about 25 percentage points. Though, due to the problems with low sample size, this difference is only significant at the 0.1 level of significance and the result should be interpreted with caution.

It is worth noting that anonymously held opinions are plausibly more predictive of anonymous acts (e.g. voting). Therefore, the notably more similar perspective shared by FrP and the H on the issue of immigration, when sentiment is covertly expressed, might have quite important political and social implications. This would be particularly important to take into account when preferences are anonymously expressed (e.g. elections, referenda). The H and FrP have been in a stable government that has collaboratively pursued more restrictive immigration policies. If the preferences of large parts of supporters of H, albeit only expressed covertly, are in-line with a more restrictive posture toward Muslim newcomers – refugee or otherwise, the result is a divergence of attitudes that is limited to the public sphere (i.e. overt expression) and the underlying acceptance of seemingly divergent policy positions should be expected. In addition, future work should have a clear prediction in other, neighbouring contexts (e.g. Denmark) where a similar pattern of cooperation between overtly incompatible political parties can be better understood if covert sentiment is taken into account.

No potential conflict of interest was reported by the authors.

Mathew J. Creighton is an Associate Professor of Sociology at University College Dublin in Dublin, Ireland. His main areas of research are immigration, education and health. He is a country co-coordinator of the European Social Survey in Ireland and has designed survey experiments in many countries and contexts. His work has been published in Social Forces, Social Problems, Social Science Research, JEMS, IMR and Demography.

Zan Strabac is a Professor of Sociology at the Norwegian University of Science and Technology in Trondheim, Norway. His main areas of research are quantitative research methods, migrations and ethnic relations and peace studies. His articles have appeared in Ethnic and Racial Studies, Sociological quarterly, Social Science Research, Politics and Religion, Journal of Ethnic and Migration Studies and Südosteuropa.