Nativism, civic nationalism and the malleability of voter attitudes

Nativism is one of the defining phenomena of the contemporary era. Yet, we know little about how malleable citizen attitudes associated with nativism and nationalism are to priming effects when media frames which deal with key issues such as immigration are introduced. In this article, we present the findings from a survey experiment fielded to a nationally representative sample of voters in Australia in May 2019. In it, we explore whether the attitudes of voters for different political parties can be primed by introducing two contrasting media frames to measure these effects. We find positive and negative frames have no effect on the attitudes of voters for Australia’s populist radical right party, but that the former has an effect on centre-right voters in Australia. Such findings have important implications for our understanding of political communication and the malleability of political attitudes.

Increasingly, social scientists are using experimental approaches to test how malleable political attitudes are. This diverse literature on framing, priming and attitude activation crosses disciplines including political science, political communication, psychology, and beyond (Scheufele and Tewksbury 2007). While the range of topics and contexts in which these studies have been undertaken is impressive in both breadth and scope, nationalism, and attitudes towards issues commonly associated with it, such as immigration, have received particular attention, which is unsurprising given their salience in contemporary politics. However, many of these studies have used party or elite frames to experiment with the malleability of political attitudes (Brader et al. 2008;Hainmueller and Hopkins 2015;Adida et al. 2016). Priming studies which use media frames, on the other hand, have received far less attention. Work on whether such interventions affect voters for different parties in different ways is also lacking. This is an especially significant question, given the argument that nativism and the radical right are now mainstream in many Western countries (Bonikowski et al. 2019;Bonikowski 2017).
Whilst the literature on frames is significant and useful for our work here (Morrison 2019; Kneafsey and Regan 2020;Tversky and Kahneman 1981), our study focusses specifically on priming. In the seminal work of Iyengar and Kinder (1987, p. 63), priming occurs where there are 'changes in the standards that people use to make political evaluations', while Scheufele and Tewksbury (2007, p. 11) suggest priming is 'often understood as an extension of agenda setting…By making some issues more salient in people's mind (agenda setting), mass media can also shape the considerations that people take into account when making judgments about political candidates or issues (priming)'. Our study tests the malleability of nationalist attitudes in Australia in response to priming through a survey experiment embedded in a nationally representative sample of voters fielded in 2019. In particular, it uses priming to investigate the role media frames play in shaping attitudes towards a salient issue oft-associated with nationalism-specifically the issue of immigrationand considers how these media frames affect voters for different parties in varying ways.
It contributes in four distinct and novel ways. First, contrary to many previous studies, our study tests the effectiveness of competing frames-one positive and one negative-about immigration. We compare the two, not only in terms of their overall effects on nationalist attitudes, but how they differently affect voters for different party types. Second, rather than artificial treatments, our study uses short videos that were originally aired on television and then widely shared on social media in 2019 as treatments. Thus, while the experiment is quite evidently artificial (as all experiments are), the treatments are not. Third, it splits nationalist attitudes into two types-nativism and civic nationalism-and seeks to examine how the treatment frames affect these separate types of nationalism differently-in other words, we are not only interested in whether, say, a positive frame decreases nativist attitudes, but if it also increases civic nationalist attitudes at the same time. Fourth, to our knowledge, it presents the first experimental research on this topic that has yet been undertaken in the context of Australia. As a country that is lauded as a 'multicultural success story' on the one hand, yet characterized by a political culture in which nativism is heavily mainstreamed on the other (Goot and Watson 2010;Mondon 2013Mondon , 2012, Australia is an extremely useful and suitable case study, with strong external validity both empirically and theoretically, for studying the effect of both positive and negative treatments in regards to immigration, and their effects on nationalist attitudes.
The article proceeds as follows. First, we outline our theory and hypotheses. Drawing on work from the literature on different forms of nationalism, we distinguish between nativism and civic nationalism, and we consider how these overlap with attitudes towards immigration. We hypothesize about the effects of proimmigration versus anti-immigration treatments on the attitudes of different types of voters. Second, we describe our case selection, data, and methods. We explain why Australia is a useful case study, our treatments, and how we developed our survey items. Third, we offer our analysis by testing the general treatment effect of our experimental study-in which we find only the pro-immigration treatment has any effect, specifically decreasing nativist attitudes and increasing civic nationalist attitudes. We then focus on the differences between the effects on different party supporters. We find that centre-right voters react to positive messages about immigration, while in contrast there is little effect on supporters for other parties-including populist radical right parties-which has important implications for how we think about such framing. We end with our conclusions, situating our findings in the literature and considering both the limitations of the study as well as potential paths forward opened by our analysis.

Theory and hypotheses
It is widely accepted that voters respond to various cues-such as policy positions and public statements-from the media, parties they support and political elites, and that this shapes their attitudes towards salient issues (Druckman et al. 2013;Hellwig and Kweon 2016;Slothuus 2016). Issues associated with nationalism are no different, and there is now a significant literature which has analyzed how different frames shape public opinion towards, among other issues, immigration (Bohman 2011;Harteveld et al. 2017). While much of this literature focusses on the priming effects of party and elite frames, some studies have sought to explore the role the media plays in increasing the salience of certain issues. One example of this comes from Sheets et al. (2016) who explored how media frames interact with party cues. In particular, they sought to determine whether certain media frames had the potential to increase support for populist radical right (PRR) parties and to shape a range of associated attitudes, including negative attitudes towards immigration.
Our approach is different. Rather than seeking to understand the role of parties in shaping public opinion (Brader et al. 2008;Sniderman 2000), or how media frames affect support for PRR parties, our approach is to explore how media frames prime attitudes of partisans to issues related to nationalism. We do so with the goal of determining whether attitudes to nationalism are more or less malleable for voters of different parties. To explore this, we separate between two conceptually distinct types of nationalism: nativism and civic nationalism. While there is some debate about the distinctiveness of these forms of nationalism (Brubaker 1999;Wright et al. 2012), we suggest that distinguishing between the two is conceptually and empirically useful as it allows us to consider how nationalism can broadly take both exclusive and inclusive forms. 1 Moreover, it follows recent work that shows that people can hold combinations of nationalist attitudes in different patterns (Bonikowski and DiMaggio 2016). 2 On the supply side, there is also a body of work that has shown that populist radical right parties have been successful in discursively combining appeals to nativism and civic nationalism as a way of making their political programs more 'acceptable' in mainstream discourse (Halikiopoulou et al. 2013;Froio 2018), meaning it is useful to look at them both in concert.
We argue that nativism should be understood as an exclusionary form of nationalism, or as Mudde (2007) describes it, as 'xenophobic nationalism'. It clearly demarcates society into 'in' and 'out' groups, and opposes people, ideas, culture, and traditions which are viewed as foreign or not reflective of the nation, however defined. In its most extreme forms, such nativism can take the form of ethic nationalism, whereby only those of a particular ethnic group (or those 'born here') are seen as legitimate citizens and members of the state (Fernandez 2013). Civic nationalism, on the other hand, is a more inclusive form of nationalism, given that it stresses shared citizenship and participation as the core feature of belonging to a nation, rather than an ethnic or cultural perception of who the 'real' members of the nation are.
While nativism can be described as illiberal given its rejection of the core values of pluralism and diversity that underlie the lived reality of migrant nations such as Australia, Canada, the United Kingdom and the United States, civic nationalism on the other hand is very much in line with liberal political traditions, given its emphasis on diversity, openness and tolerance (Laegaard 2007). As such, we suggest civic nationalism is also more congruent (at least theoretically) with voting for 'mainstream' parties in multicultural societies, given that most major parties in such societies endorse, or at very least pay lip service to, such values. On the flipside, manythough not all-of the parties who espouse nativist positions are those who are commonly classified as the 'populist radical right'. Indeed, Mudde specifically notes that nativism is 'the core feature of the ideology of this party family ' (2007, p. 26), even more than the other two ideological features, populism or authoritarianism.
The most salient issue that the different forms of nationalism are related to in contemporary politics is undoubtedly immigration (Givens and Luedtke 2005; Dennison and Geddes 2019; Nguyen 2019). Questions regarding levels of immigration, the kinds of immigration (whether skilled or non-skilled), and where immigrants should be drawn from are heavily contested in liberal democracies across the globe, with broader questions about immigration often spurring debates about questions such as: is immigration a positive or negative force for a nation-state? How do immigrants 'become' part of the nation-through assimilation or integration? Who is the impetus on in terms of how immigrants are included within the national polity? Nativist and civic nationalist attitudes inform different answers to such questions.
When it comes to elite messaging around issues of immigration in multicultural liberal democratic states like Australia, the United Kingdom and the United States, something of a paradox exists (Tavan 2007). Formally, state policy tends to support multiculturalism, relatively high levels of immigration, 3 and policies that accord with civic nationalist values around integration, openness and the social good of diversity. Yet, elite and media messaging is often quite different to this. Major parties often hedge their bets, offering at times inconsistent, contradictory, or ambiguous cues around immigration. At times, they will praise the 'success story' of multiculturalism and the positive impact immigrants have on their societies, while at others demonizing immigrants as a drain on the public purse, or framing them as a health or security threat (Bale 2008;Alonso and Fonseca 2012). In stark contrast, messaging from PRR parties (and indeed, a substantial swathe of tabloid media in such countries) is unambiguous: it is nativist, and they put forward a consistent negative frame when it comes to immigrants coming from regions, religions or cultures perceived to be incompatible with the natives (Vliegenthart and Roggeband 2007;Helbling 2014;Benson 2013).
This leads us to the question: how do voters for different parties react to positive or negative framing of immigration? Are nativist attitudes or civic nationalist attitudes increased or decreased by being exposed to such frames? And are some voters' attitudes more malleable to effects than others in this regard? As outlined in Table 1 below, we theorize that a range of effects are likely for different sets of attitudes associated with nationalism when positive versus negative frames are introduced. In short, we expect that a positive frame will have a larger range of effects on voters. We suggest it will decrease the nativist attitudes and increase the civic nationalist attitudes of these voters. We also theorize that the negative frame will decrease the civic nationalist attitudes of voters. We outline our reasoning for our theorizing below.
First, as alluded to earlier, negative frames about immigration, refugees and associated issues' effects-whether economically, or in terms of security or cultural identity-on Australian society are extremely common. Indeed, a number of scholars have argued that nativism is ostensibly mainstream in Australia (Goot and Watson 2010;Mondon 2013Mondon , 2012. Given the dominance of this frame, we suggest that introducing yet another negative frame will have little priming effect on the exclusionary attitudes associated with nativism for voters. Our first hypothesis therefore is:

H1
The anti-immigration frame will have no effect on the nativist attitudes of voters However, we also suggest that given the ubiquity of nativist messaging in Australia (Stokes 2000;Wear 2008;Kefford and Ratcliff 2021), a pro-immigration frame which highlights the significant benefits of immigration and which presents migrants in a positive light will likely have an effect. While at the individual level, avoiding cognitive dissonance would normally lead us to predict that pro-immigration frames fall on deaf ears with nativist voters, the Australian case provides a unique social configuration where anti-immigration frames dominate the public debate within a country that has some of the highest rates of migrants of any advanced democracy. Thus, pro-immigration frames contradict this general trend, which might make them more impactful than often-repeated anti-immigration tropes. Our second hypothesis therefore is: H2 The pro-immigration frame will weaken the nativist attitudes of voters In contrast to the dominant frame of nativism in Australian political discourse, we argue that civic nationalism is more ambiguous. As noted earlier, while simultaneously promoting Australia's punitive asylum seeker policies and notoriously tough immigration policies, the major political parties are also prone to talking of the benefits that migrants bring-socially, culturally, and economically-to the nation. This appears contradictory, but we suggest that there is likely an important distinction between those who primarily hold nativist attitudes as opposed to those who primarily hold civic nationalist attitudes, and that this ambiguity will manifest in how attitudes to civic nationalism are affected (and primed) by competing frames. Indeed, we argue that the negative frame will decrease the more inclusive civic nationalist attitudes amongst these voters, while a positive frame will have the opposite effect. Our third and fourth hypotheses are: H3 The anti-immigration frame will weaken the civic nationalist attitudes of voters H4 The pro-immigration frame will strengthen the civic nationalist attitudes of voters

Case selection, data and survey items
Australia is a suitable case study for studying the effect of both positive and negative treatments about immigration for several reasons. First, Australia is widely noted as a multicultural 'success story', with over a quarter of the population being born overseas, and half of all households having at least one parent being born overseas. On the flipside, Australia also has a long history of racism and exclusionary nationalism that can be traced to the pre-federation period (Blackton 1958) and the first piece of legislation the newly federated Commonwealth of Australia passed in 1901 was the so-called 'White Australia Policy' (Fozdar et al. 2015). According to Stokes (2004, pp. 8-9), "White Australia contributed both to the consolidation of an Australian racial identity and to the exclusion of particular outsiders. But racial identity was not just maintained by external programs that excluded immigrants; it also operated internally through the exclusion and subordination of the Indigenous inhabitants of Australia". Hence, nativism is strongly embedded in Australian political culture; and the country has some of the most punitive policies on immigration and refugees across advanced democracies, which enjoy wide political support.
This somewhat contradictory situation makes it ripe for analysis in terms of how nativist and civic nationalist attitudes operate and makes it a 'most likely' case amongst similar multicultural Western countries such as the United States and the United Kingdom, the latter of which has looked towards Australia in recent years in terms of its immigration model. Second, as mentioned, Australia has not been the site of experimental research on this topic as yet (to our knowledge). With this in mind, we argue that the case of Australia is externally valid for both empirical and theoretical reasons. Empirically, Australia's combination of a diverse populace together with a party landscape in which a PRR party 4 has had both relatively longlasting electoral success and agenda-setting success in terms of shifting the policy positions of the centre-right (Mondon 2013) means that it is comparable to several Western European countries in similar situations (Van Spanje 2018; Akkerman et al. 2016). Pauline Hanson's limited but long-lasting success, can be seen both as a product of the demand for nativism in Australia (Kefford and Ratcliff 2021), and her ability to adjust who was the focus of her exclusionary nationalism. Initially, as was evident in Hanson's maiden speech to the Commonwealth parliament in 1996 this was Asians, claiming "…we are in danger of being swamped by Asians…They have their own culture and religion, form ghettos and do not assimilate" (SMH 2016). Since her return to the Senate in 2016 the focus of her exclusionary nativism has, primarily, become Muslims. Theoretically, the combination of different types of nationalist attitudes that we find in Australia are not unique to the country, but also exist in other national contexts (Simonsen and Bonikowski 2020;Bonikowski and DiMaggio 2016) and as such, findings about the malleability of such attitudes for voters should be theoretically transferable to and testable in other cases.
When it comes to the question of how to measure nationalist attitudes in surveys, there is a long tradition in political psychology and voting behavior studies in distinguishing between civic or liberal nationalism and ethnic/cultural nationalism (Bonikowski 2017; Bonikowski and DiMaggio 2016). We therefore have included questions into our survey (see below) that seek to capture whether our frames prime one or both 'types' of nationalism. In the first case, the literature on ethnic and cultural nationalism is extensive, and has served as the basis for the theoretical and empirical literature on nativism and nativist attitudes. While such attitudes are commonly associated with anti-immigrant attitudes, which we seek to capture with questions about immigrants keeping their own languages and customs, the 'loss' of Australian culture, and the effect of immigrants on Australian society; we also acknowledge that the out-group to the 'natives' in-group can of course include anyone else perceived to be a threat to the physical or existential security of the 'nation'. In terms of ideas, nativism often relates to notions of 'foreign' cultures or traditions that have been introduced by migrants. We have thus designed items which measure attitudes to each of these underlying dimensions of nativism in Australia: race, immigration, and identity.
In the literature on liberal or civic nationalism, pride in one's country and institutions, patriotism, and a subjective sense of feeling attached to the nation are some of the key dimensions used to measure the extent of these attitudes (Bonikowski 2017;Bonikowski and DiMaggio 2016;Arts and Halman 2005). We have therefore used question items which attempt to tap attitudes to these dimensions, and which operationalize an inclusionary form of nationalism. As per Table 2 below, this includes questions about respecting Australian laws and institutions, the importance of 'feeling' Australian, and whether one has to live in Australia for a long time to 'be' Australian. These questions and this subtype of nationalism contrasts starkly with the underlying dimensions of nativism. 5 To investigate how different frames about immigration prime the attitudes of different voters, we randomly sorted our respondents into three groups. Our control group was only asked for their attitudes to nativism and civic nationalism. The other two groups were shown a video clip before these questions. As treatments, we use two videos broadcast on television in Australia that were also shared widely on social media, thus accounting for the type of hybrid media that can span across the spheres of legacy and digital media (Jungherr et al. 2020), the latter of which has become increasingly important for PRR parties' communication strategies (Jungherr et al. 2019). Experiments of this kind have become increasingly common in the social sciences. While there are reasonable concerns about external validity in any experiment (Barabas and Jerit 2010), the videos we use as our treatments are 'real' videos, and as such, this enhances the external validity of our findings.
The first of these short videos, the negative frame, was broadcast on the Channel 7 network, a commercial station in Australia, and was then spread across social media via a range of actors, including the leader of Australia's PRR party, Pauline Hanson (Hanson 2017). According to the statistics on Hanson's Facebook page, the video has been viewed on that page alone over 327,000 times. It frames immigrants as incompatible with 'Australian values' and the 'Australian way of life'. Indeed, it features an interview with Hanson, who says that immigrants 'will impose their beliefs, their way of life, and it will undermine our society'. This video is representative of a great deal of the widespread anti-immigrant media content in Australia. It also aligns very clearly with the idea of nativism.
The second video, the positive frame, was broadcast on television in Australia on the SBS network, the publicly funded multicultural broadcaster. On the SBS Facebook page alone, this video has been viewed 146,000 times (SBS 2018). It shows refugees finding a home in regional Australia. Both the refugees and 'locals' speak positively about the way that refugees have become an integral part of the local community, thus pushing against the cultural threat myth-that Australian culture and values will be lost or diluted due to immigration and that immigrants will seek to impose their values on society. We contend that this framing is one that, theoretically at least, should align with and strengthen civic nationalist attitudes, as it shows that refugees have had a positive effect, and that they have become an important part of Australian society. We fielded our experiment as part of the Australian Co-operative Election Study 6 that also contained questions regarding a range of demographic characteristics as well as respondents' political preferences and behaviors. This survey was fielded in May 2019 through YouGov Australia, who provided a non-probability sample based on their panel of volunteer survey respondents. Our sample of 1049 respondents was stratified by gender, age, and region. Within the two experimental groups that were treated with the videos, about 90% of respondents reported to not be familiar with the specific video (positive frame video: 95%, negative frame video: 90%). Thus, while the videos used are realistic treatments, we can assume that the specific videos provide a novel frame for respondents.
To better understand what effect these treatments would have on voter attitudes, we compare the likelihood to agree with the question items outlined in Table 1 between those respondents who watched the specific videos and our control group who received no treatment but were only asked the nativism and civic nationalism questions. To measure the effect of the treatments, we first investigate the average treatment effect on the treated (ATT), which calculates the causal effect among those who get the treatment, to test for differences between these groups (Ho et al. 2007, p. 204). Furthermore, we use regression analysis as a robustness check (Marx and Schumacher 2016). Both analyses include control variables that are commonly used to analyze political attitudes (gender, age, income, education, residence). In a second step of the analysis, we then split our sample by party supporters and repeat both analysis steps for each group separately to test our four hypotheses. To distinguish these voters, we use the respondents' answer about how they would vote in the 2019 Australian federal election. This means we can separate results for those who indicated how they would vote in the 2019 federal election for the Australian Labor Party (Labor), the Liberal Party of Australia, the Nationals, the Greens and One Nation. 7

Analysis: how malleable are Australians' nativist and civic nationalist attitudes?
Before turning to the effect of the experimental treatments on the Australian respondents' attitudes for nativism and civic nationalism, we need to briefly set the benchmark for both of these variables. Both nativism and civic nationalism are measured as continuous variables ranging from 0 (no nativism and no civic nationalism) to 3 (very strong nativism and civic nationalism) by combining the respective survey items from Table 2. 8 Amongst our respondents, civic nationalism was slightly stronger, with an average score of 1.9 (SD 0.5), than nativism, with an average of 1.5 (SD 0.6). Figure 1 shows the distribution of both variables as violin plots, revealing not only the higher level of civic nationalism but also that also a higher level of agreement among the respondents in this variable compared to the greater dispersion in nativism.
For a first test as to whether our experimental treatments had any effect at all, we calculate the average treatment effects of the treated (ATT) for all respondents together. 9 Due to missing values, our analysis is restricted to 958 respondents. Table 2 shows the mean values on nativism and civic nationalism in the control group and the average treatment effect of the two frames. It shows that, overall, only the video with a positive frame had an effect. As predicted, watching a video that positively frames immigration reduces overall nativism and increases civic nationalism. On the other hand, showing Australians another video that paints immigrants in a negative light does not seem to sway respondents on either attitudinal dimension. Given that Australian history and political culture is steeped with nationalist and  8 We also checked that the items proposed in Table 2 indeed form an attitudinal dimension each by running iterated principal-factor analyses. The results are shown in Table 8 in the appendix and confirm that these items form the two individual factors. While some items do not load strongly onto the two factors, all items are retained as the composite dimensions are primarily constructed on theoretical grounds. Table 7 shows the experimental treatment effects for the individual nativism and civic nationalism items. Furthermore, we add a robustness check combining only the two strongest loading items for civic nationalism and re-estimating the main models shown in Tables 3 and 4. Tables 9 and 10 show that our analysis is robust against this change in the dependent variable. 9 We used the specialized STATA command teffects with the regression adjustment (ra) subcommand to account for the control variables and the atet option to obtain the average treatment effect of the treated.

Table 4
Average treatment effect of the treated by vote intention N = 795 Robust standard error in parenthesis ***p < 0.001, **p < 0.01, *p < 0.05; controlling for age, gender, income, education and urban/rural

Different effects for different party voters?
We then split our sample by who voters suggested they would vote for in the 2019 federal election. For our purposes, we focus on respondents who indicated they were likely to support the parties that made up the centre-right coalition government at the time the survey was conducted (the Liberal Party of Australia and the Nationals), the centre-left opposition (the Australian Labor Party), and the other two most significant minor parties, the Greens and Pauline Hanson's One Nation. By excluding voters of other minor parties, Independents and individual candidates, the sample is slightly reduced to 795 respondents. The results in Table 4 show several interesting patterns. First, when we look at the average nativism and civic nationalism in the control groups, we find that the centre-right Coalition parties, the centre-left ALP and the PRR party, One Nation, have higher values than Greens voters on both dimensions. While this is not surprising for nativism, the difference on the civic nationalism dimension was not necessarily expected. Second, while the mean value for civic nationalism is higher than for nativism among Coalition, ALP and Greens voters, we find the opposite among One Nation voters. However, the mean value for civic nationalism for One Nation voters is still the highest of any group of voters: in short, One Nation voters are both nativist and civic nationalist. Turning to our hypotheses, Table 4 confirms our expectations expressed in hypothesis 1 that the negative frame treatment has no effect on the nativism of any group of voters. Contrary to hypothesis 3, the analysis also shows that this negative frame has no effect on civic nationalist attitudes for any group of voters as well. That this negative frame does not affect nativism or civic nationalism for any group of voters indicates that Australian respondents are potentially so used to negative reporting about immigrants and immigration that further framing in this direction makes no difference, no matter their party preference.
When we investigate the effect of the positive treatment, meaning the pro-immigration frame, we find confirmation for hypothesis 2. While this treatment did not affect all groups of voters in the same way, it decreased nativism most clearly among Coalition voters. The positive frame also had a small effect on Coalition voters' civic nationalist attitudes, thereby providing partial confirmation of hypothesis 4. These two effects of the positive frame on voters for the major parties shows that these voters' attitudes towards immigration are indeed malleable. For One Nation voters, on the other hand, the ideological foundation of these voters on nationalist issues-whether nativist or civic-seems to be so strong that no new information affects their attitudes.
Based on the OLS regression explaining nativism and civic nationalism (see Table 5 in the appendix) respectively, Figs. 2 and 3 show the marginal effects of the two treatments and confirm the findings from Table 4. Figure 2 shows the effect of the two treatments (each compared to the control group) for the different voter groups. Only for Coalition voters do the treatments have an effect, albeit only the positive framing of immigration that reduces the level of nativism. Since the positive frame has no effect on One Nation voters, we find overall evidence for our hypothesis 2. As the negative frame has no effect on the nativism of either voter group, we also find evidence for hypothesis 1.
Turning to hypotheses 3 and 4, Fig. 3 shows the marginal effects of the two treatments on civic nationalism in the different voter groups. This also confirms our findings from Table 4 in that the only significant treatment effect is the positive frame among Coalition voters, which increases civic nationalism attitudes.

Conclusion
Nativism is mainstream in Australia. Voters for the centre-right Coalition have expressed time and time again a set of policy preferences on immigration and questions of nationalism that are arguably congruent with that of many PRR voters internationally (AES 2020). 10 This includes long-running support from the majority of Coalition voters on policies such as 'turning back boats' containing asylum seekers. Voters for the centre-left Labor Party, have a less radical stance, but still around a third Fig. 3 Treatment effect on Civic Nationalism. Baseline = control group 10 As an example of this, longitudinal data from the Australian Election Study shows that in the seven Australian federal elections since 2001, at least 63% of respondents who said they voted for the Liberal Party also said boats carrying asylum seekers should be turned back. Respondents who said they voted for the Nationals had similar levels of agreement. For those who said they voted for the Labor Party, the number agreeing with this statement has declined over time, from a high of 53% in 2001, to 37, 35 and 38% over the last three elections respectively. Source: (AES 2020). of voters consistently support policies which are, for example, tough on migrants and refugees. To investigate whether we could prime voters for the various Australian parties, we have presented the findings of an original survey experiment which aimed to explore the malleability of political attitudes associated with two forms of nationalism-nativism and civic nationalism. In doing so, our goal was to ascertain whether the attitudes of voters are likely to more malleable when primed with positive or negative media frames about immigration. We have demonstrated that the larger effects were almost entirely amongst those voters who expressed their support for the centreright parties. This is a significant finding: on the one hand it supports previous research which suggests that parties are important in shaping political attitudes of partisans. On the other hand, it shows that media frames can, at least momentarily, override party framing as our treatment with a positive media report about immigration made centreright voters less nativist, even though their party tows a strong and consistent nativist line. Our capacity to produce small but statistically significant priming effects in the nativist and civic nationalist attitudes-a negative effect in the former case, and a positive effect in the latter-of these voters indicates that the ubiquity of nativist discourse and messaging (and its subsequent effects on attitudes) has not necessarily contributed to a hardening of the attitudes on these issues. 11 Moreover, while a small sample and needing to remain cautious about extrapolating from such a number, One Nation voters were almost completely unaffected by either frame, indicating that if media, party or elites hope to influence PRR voters' views, new thinking may be needed in terms of how to frame communication towards them.
There are inevitable limitations of any study of this type. First, the items we have used to tap nativist and civic nationalist attitudes are, like any attitudinal measure, open to revision, and may require modification to be used in other cultural contexts. Moreover, as mentioned earlier, nationalist and civic nationalist attitudes, while conceptually different, are not necessarily opposed to one another: one can hold nativist views as well as civic nationalist views, and there are different configurations of nationalist attitudes in the electorate that go beyond the limits of voters for certain parties, as the work of Bonikowski and DiMaggio (2016) has shown. Future work would do well to examine how these configurations operate, especially outside the US where most of the work has thus far been undertaken. Second, the single case used here means that the generalizability of our findings will need to be tested in other advanced democracies. Third, it is of course likely that the effects of individual frames will decay rapidly, as is expected of all forms of political communication. Hence, it is likely that such frames would need to be layered and reinforced to have a lasting effect on public opinion and political behavior. Additionally, there are temporal limitations to these findings. We conducted one survey with a particular set of contextual conditions. Evidently, panel studies would provide a stronger set of findings. Fourth, a limitation and one potential explanation of the small effects are the media selected. Arguably, more provocative, organically generated content created by, for example, a PRR party like Hanson's and another promoting an alternative version of nationalism may have led to a different set of findings. This is a limitation that could be addressed in future studies. Nonetheless, given how salient immigration is in most advanced democracies and the rise of parties that play on nativist attitudes globally, the empirical and theoretical findings presented here are significant for scholars interested in political attitudes, priming, and supply-side analyses. Future work could extend on our findings by employing panel studies to increase our understanding of these effects and how quickly they decay over time. At a time in which nativism is on the rise in many advanced democracies (Mudde 2019), work that explores how such attitudes can potentially be decreased is vital, and our study presents an important and original contribution to the scholarship in this regard.

Appendix to Nativism, Civic Nationalism and the Malleability of Mainstream and Populist Radical Right Voter Attitudes
(1) Further information about main steps of analysis See Tables 5 and 6.     Table 8 Unrestricted iterated principal factor analysis for nativism and civic nationalism As the theoretically argued civic nationalism dimension does not empirically cluster as expected, we conduct a robustness check. First, we create a second civic nationalism variable averaging only the first two items (Respecting Australian political institutions and laws; Importance of Feeling Australian) that cluster highly on Factor 2 in Table 8. This new variable correlates strongly with the civic nationalism variable used in the main analysis (corr = 0.76, p < 0.000). We then repeat the above analysis using this new dependent variable and comparing the effects with the civic nationalism variable used in the main article. Tables 9 and 10 replicated the analyses in Tables 3 and 4 in the main article, respectively. The comparison shows that the treatment effects on the 2-item measure of civic nationalism is slightly stronger than in the main analysis but significance and direction of the effects are confirmed

Table 10
Average treatment effect of the treated among mainstream and PRR voters, comparison of different item combinations for civic nationalism, N = 795 Robust standard error in parenthesis ***p < 0.001, **p < 0.01, *p < 0.05; controlling for age, gender, income, education and urban/rural