Fueling the Populist Divide: Nativist and Cosmopolitan Preferences for Representation at the Elite and Mass Level

Although populist leaders often employ an anti-elite discourse which presents the elite as unable or unwilling to represent ordinary citizens, we know very little about who elites actually think should be represented, or how this differs, if at, all from what ordinary citizens want. In this article we find that there is a considerable difference between the groups that voters want to see represented in parliament and those which political elites want to see represented. In particular, we find that political elites tend to hold far more ‘cosmopolitan’ preferences than ordinary voters, and prioritize the representation of greater diversity in parliament based on the groups politicised by the new social movements and identity politics of the 60s and 70s, such as women, ethnic minorities, LGBT and the disabled. By contrast, voters more often hold nativist preferences than political elites and more often prioritize the representation of groups such as the working class, and white local people. Moreover, British voters who hold nativist preferences of political representation are more likely to be politically alienated and more likely to support Brexit.


Introduction
In recent years there has been a dramatic increase in support for populist parties around the globe (Inglehart & Norris, 2017). As a consequence, anti-elite discourse is now a central feature of public debate and arguments about who our elites are, and who they should represent, dominate the airwaves. Many populist politicians frame themselves as outsiders from the political class, even if this does not accurately reflect their actual association with the political establishment (Barr, 2009). In doing so they seek to cultivate an 'us vs them' dichotomy (Mudde, 2007, p. 63), which presents the elite as unable or unwilling to represent ordinary citizens (Ignazi, 1996;Scarrow, 1996;Schedler, 1996). Populists therefore seek to speak for the 'silent majority' of 'ordinary, decent people' whose interests and opinions they claim are ignored by arrogant elites, corrupt politicians, and strident minorities (Canovan, 1999, p. 5). This construction of "the people" is facilitated by the invocation of the people's enemies, and the "people" often come to know who they are by who they are not (Oliver & Rahn, 2016). Consequently, nativism and racism are common in populist appeals, particularly in democracies facing immigration pressure. Yet despite the common charge that the political class are out of touch and unrepresentative of ordinary people, we currently know very little about who political elites think should be represented or how this differs, if at all, from ordinary citizens. Using British data collected in 2015 at both the elite and mass level, we investigate which social groups political elites, operationalized as national electoral candidates and incumbent legislators, think should be represented in the British House of Commons, and whether this differs from the groups that ordinary citizens would like to see represented.
We develop a framework for analysis which challenges the idea that people view descriptive representation in purely egocentric terms-and just want to see more people like themselves in their legislatures. We contend that support for the representation of different groups may also take a more ideological form. That is people may have distinctive views on what they think society should look like-and consequently which parts of society should be better represented politically. And whereas it might be assumed that voters simply want their legislature to be a reflection of the diversity of society at large, and so have a preference for the increased representation of all marginalized groups, we argue that this preference is not universally held and represents a distinct 'cosmopolitan' value orientation about how society should be organised and hence what legislatures should look like. To this end, we find that support for increasing the representation of different social groups, at its poles, reflects two competing visions of society: cosmopolitan vs nativist.
The core idea of nativism is that a state should comprise 'natives', and 'nonnatives' are to be treated with hostility (Mudde, 2007, p. 138). Accordingly the interests of native-born or long established communities should come before those of immigrants (Pappas, 2016, p. 123) or their descendants. This ethnically based conception of citizenship has little rational basis in fact given all human populations share the same common ancestors and, in the UK, all modern waves of immigration have been encouraged by the state rather than enforced on the nation via colonial conflict. 1 Nonetheless there exists a nativist ideology that imagines a white British nation, and challenges the legitimacy of black and minority ethnic immigrants and their descendants' citizenship. 'Nativist' attitudes towards representation fall on a spectrum, but overall, we expect nativists to prioritize the representation of ethnically exclusive-usually majority-groups. In the UK white nativists claim to promote the interests of the white Britons and 'local' people, rather than ethnic minorities and members of non-Christian religious communities, currently particularly Muslims (Ogan et al., 2014).
By contrast cosmopolitanism derives from the Greek word meaning 'a citizen of the world' and stresses that all human beings belong to a single community and emphasizes the equality, commonality and interconnectedness of all human beings irrespective of geography. 'Thick' notions of cosmopolitanism move beyond geography to consider other social inequalities. Thus "all moral principles must be justified by showing that they give equal weight to the claims of everyone, which means that they must … be directly universal in their scope" (Miller, 1998, p. 166). 'Cosmopolitan' attitudes towards representation therefore prioritize equality of representation for all people and promote greater diversity than is currently embodied in historically homogenous political institutions. Thus a cosmopolitan approach to representation prioritises all groups that are under-represented relative to their proportion in the population, including working class communities, alongside members of those groups with a history of discrimination, that were at the forefront of the new social movements of the 1960s and 1970s, such as women, ethnic minorities, LGBT and the disabled.
We show therefore that there is disagreement regarding which groups merit greater political representation, and we find evidence of a stark divide between political elites and ordinary voters. In the UK political elites are overwhelmingly likely to favour cosmopolitan representation, and are much more likely to do so than ordinary voters. By contrast, voters more often hold 'nativist' preferences of representation than political elites. There is thus evidence of a 'representation gap' between those groups that political elites think should be represented and those groups that voters think should be. This can create a fertile breeding ground for populism (Oliver & Rahn, 2016, p. 194). When a 'representation gap' exists between political representatives and large sections of the electorate people can feel that they are not being (well) represented by the elites in power, and, accordingly, will criticize-and even rebel against-the political establishment" (Mudde & Kaltwasser, 2017, p. 10). This could set the stage for a populist struggle "to give government back to the people" (ibid.) or "take back control." Consistent with these expectations we find that voters who favour nativist representation are more likely to exhibit populist attitudes and behavior than those who support the status quo.
The case of Britain provides a valuable and potentially insightful example for exploring these disjunctures between elite and mass preferences for political representation. The tension between demands for better political representation for the social groups that gained more prominence during the 1960s and 1970s such women, ethnic minorities, and LGBTs and white working class [men] has become increasingly fraught in the last decade in many western democracies including the United States, Denmark, Sweden, Italy, Germany, the Netherlands and Britain. Britain, in common with many other nations, has seen a surge in support for right-wing populist movements and a reconfiguration of electoral politics around cultural issues and questions of identity. These 'culture wars' each have their national context but share some drivers including the impact of globalization on working class western communities, the atomization of society resulting from contemporary neoliberalism and cultural backlash to the changing social values and norms resulting from the successes of the new social movements of the 1960s and 1970s (Ignazi, 1992;Mudde & Kaltwasser, 2018).
As in many other European democracies, the focus for much of the populist mobilization in Britain has been the EU, with populist parties portraying "the EU as a project that threatens the sovereignty of the native people and, through the opening of borders, the cultural homogeneity of nations" (van Kessel et al., 2020, p. 67). In the UK the impact of this populist anti-EU sentiment has been profound with the success of UKIP pressuring the Conservative party to promise and then to hold a referendum on Britain's withdrawal from the EU; a seismic event which has transformed British politics. The UK's deployment of a referendum to address the question of Britain's membership in the EU, rather than dealing with the issue through the processes of representative democracy, is a striking example of the success of populist direct appeals to the people, making the UK an excellent test case for exploring the tensions underlying contemporary Western political and party systems more generally.
The policy failures of mainstream parties to meet the needs of the 'left behind' communities of post-industrial societies have been well documented but considerably less attention has been paid to their descriptive representation and how this may have fueled populist sentiments. While we cannot assess whether a divergence between the preferences for group representation among the political elite and the mass public have played a role in fueling populist anti-elite sentiment, we can explore the context in which this mobilization took place. In common with many older democracies, over the last forty years or so the UK has witnessed substantial changes in the social groups that established political parties represent. In particular, the proportion of working-class politicians in the British parliament has sharply declined: around 40% of Labour MPs were working class in 1951 compared to less than 10% in 2017. During the same time, other social identities have attracted greater political attention, particularly those identities that have been the basis of discrimination, such as gender, sexuality, ethnicity and disability.
In Britain, and elsewhere, these identities have been the subject of considerable legislation aimed at promoting social equality, and have also been the subject of debate within political parties about the extent to which they represent these forms of diversity. In particular, the proportion of women and ethnic minority politicians in the British parliament has increased. In 1979 just 3% of MPs were women compared to 34% in 2019. Likewise, the number of black and minority ethnic MPs has increased from zero in 1979 to 65 (10%) in 2017 (compared to 14% in the population).
The representation of these different groups need not be zero sum-and in theory there is no reason why the increased representation of women and ethnic minorities should be at the expense of the representation of working class MPs-especially given that women and ethnic minorities are disproportionately over-represented Political Behavior (2021) 43:1707-1729 among those living in poverty in the UK. 2 However, the changes that have occurred, with an increase in the representation of protected groups who have faced discrimination, taking place alongside the erosion of white men with working-class occupational backgrounds might signal to some that in practice the representation of these groups is in fact zero sum. If this is the case, we may then expect demands for group representation to follow these two competing trajectories: cosmopolitan on the one hand, and nativist on the other.
To examine this possibility this article proceeds in three parts. First, we show how our framework for studying attitudes towards representation builds on and develops past research. Second, we show how voters and candidates prioritize the representation of different social groups, and document the gap between them that exists with respect to nativism and cosmopolitanism. Third, we examine the attitudinal and behavioural correlates of these attitudes towards group representation. In the conclusion we reflect on the implications of our findings and their generalizability to other contexts.

Descriptive Representation
We examine what drives support for the political representation of different social groups. We ask: 'to what extent do voters and candidates differ in their preferences for representation?', And 'to what extent do voters who hold different preferences for representation differ in their political attitudes and behaviour?'.
Descriptive representation occurs when the representative stands for a social group by virtue of sharing similar characteristics. According to one influential perspective sociodemographic dissimilarity with a political figure (e.g., party leader) tends to decrease a voter's expected utility from the election of that person (Cutler, 2002). Thus, from an identity politics perspective, women should prefer women candidates, black people prefer black candidates and so on. Part of the rationale for this argument is that in the eyes of the voter there is a link between descriptive and substantive representation; where, for example, women are thought to be more likely to advance policies that are in the interests of women. This line of argument treats support for the descriptive representation of different groups in egocentric terms, meaning that citizens want to see more people like themselves in legislatures. However, support for the representation of different groups may also take a more ideological form. That is people may have distinctive views on what they think society should look like-and consequently which parts of society should be better represented politically.
Preferences for the representation of different social groups may therefore go beyond a simple preference for 'more of one's own' particular identity group, and signal a wider sociotropic outlook which reflects citizens' views of what a modern legislature should look like and who it should represent. For many people, this may embrace ideas of diversity and difference, but not for all. It is this possibility that we 1 3 investigate. We examine whether attitudes towards representation prime a broader cosmopolitan identity, that embraces many of the social changes that have taken place in modern society and demands further progress, or a narrower nativist identity, that is more resistant to these changes. This leads to our first hypothesis that attitudes towards the representation of different social groups prime different conceptions of representation, which take into account the wider ideological symbolism of which groups people think should be represented.
Hypothesis 1 Support for the representation of greater diversity in parliament ("cosmopolitan representation") is empirically distinct from more ethnically exclusive support for working-class MPs who are not from ethnic minorities or of the Muslim faith in parliament ("nativist representation").
Although there is speculation as to what motivates populist movements (economic conditions, class stratification, new media technologies, etc.), one important source is to do with the view that existing political elites do not represent ordinary people (Barr, 2009). A long line of research in the US suggests that people tend to think about parties in terms of other, longer-standing groups (Ahler & Sood, 2018;Berelson et al., 1954;Campbell et al., 1960;Converse, 1964;Green et al., 2002;Hetherington & Weiler, 2009;Rothschild et al., 2019). Historically, the groups that come to mind when people think about the parties tend to be shared, with Democrats, independents, and Republicans often associating the same groups with the parties (Green et al. 2002;Rothschild et al., 2019). In the US these group-party associations tend to endure, and these shared associations reflect a tendency to think about parties in terms of prototypes-abstract composites of characteristics associated with the party, akin to Lippman's "pictures in our heads." (Ahler & Sood, 2018). For instance, when thinking about the US parties, one may call to mind a Southern, evangelical Republican or a young, non-white Democrat. Extending this analysis recent research has explored whether "parties' social composition drives partisanship" (Ahler & Sood, 2018, p. 964) and whether social sorting has occurred in the US such that individuals' group memberships and partisanship are increasingly aligned (Mason & Wronski, 2018).
However, in the UK these group-party associations have weakened markedly over time (Evans & Tilley, 2018). In particular, the association between the working class and Labour has dramatically declined, and changes in the social composition of the party has in turn influenced patterns of partisanship (Evans & Tilley, 2018;Heath, 2016, p. 18). The groups that ordinary people want parties to represent may therefore have been de-coupled from the groups that those parties traditionally did represent, and consequently may also have been de-coupled from the groups that parties now say they do want to represent. Our next hypothesis tests the extent to which a 'representation gap' exists in practice and whether political elites in Britain are indeed more cosmopolitan in their political preferences than ordinary voters, over and above any differences that exist between parties. Political Behavior (2021) 43:1707-1729 Hypothesis 2 Support for cosmopolitan (nativist) representation will be higher (lower) among candidates than voters.
Our last two hypotheses relate to the attitudinal and behavioral correlates of these ideological preferences for representation. Our third hypothesis is motivated by the intersection between nativism and populism. A now widely accepted definition of populism is Cass Mudde's description of it as a thin ideology founded on the belief that society is divided into two distinct groups 'the pure people' versus the 'corrupt elite' (Mudde, 2007, p. 23). Populism and nativism are distinct phenomenon (Iakhnis et al., 2018, p. 2) but they combine together to form the ideologies of the populist radical right who claim to represent the 'true people' and the 'true nation' and populism in contemporary advanced democracies is most commonly associated with nativist beliefs (Inglehart & Norris, 2016). The empirical link between nativist and populist values leads us to predict that support for nativist representation will be related to anti-elite attitudes, and that as a result nativist voters will not feel well-represented by the current crop of MPs in the House of Commons. However, this does not imply that we expect cosmopolitan voters to feel much better represented either. Voters with a cosmopolitan preference for political representation may report higher levels of anti-elite sentiment than those who support the status quo in parliament because, despite progress, the British House of Commons remains over-representative of white middle class men. Thus, both voters with nativist and cosmopolitan preferences for political representation may perceive a gap between the current composition of the House of Commons and their ideal distribution of group representation and thus express greater anti-elite sentiment than citizens who are satisfied with the current state of political representation in the UK.
Hypothesis 3 Voters with a nativist/cosmopolitan preference for representation are more likely than those who support the status quo to report that politicians 'don't care what people like me think'.
Relatedly, our final hypothesis is that support for nativist representation is related to support for populist movements, such as the vote to leave the European Union. There is a considerable body of research that links hostility to the EU to nativist values (Iakhnis et al., 2018;Jennings & Stoker, 2016) as attitudes towards the EU and nativism are both associated with anti-cosmopolitan or 'closed' world views. We test whether nativist preferences for political representation are also associated with support for Brexit-and whether this goes beyond simply a hostility towards immigration. On the other hand, we anticipate that voters who prefer cosmopolitan representation will be open to supranational projects such as EU membership and the associated internal migration and will be less likely to favour Brexit than those who prefer the status quo regarding political representation. Thus, we explore whether demands for Britain to exit the EU are linked to preferences for who should occupy elected office.

Hypothesis 4
Voters with a preference for nativist (cosmopolitan) representation are more (less) likely than those who favour the status quo to support Brexit.
In the next section we describe our measurement strategy and the novel data sources that we use to test these hypotheses.

Data and Measurement 3
The analysis conducted in this paper is based on two sources of data: the British Election Study (BES) 2015 Internet Panel Study Wave 6 and the Representative Audit of Britain (RAB). 4 Importantly, both of these surveys were carried out well before the UK's Referendum on EU membership took place, and well before the referendum was even called. They thus provide a valuable insight into the context in which the referendum took place. The Representative Audit of Britain is a survey of all candidates who stood for the Conservative, Green, Labour, Liberal Democrat, SNP and UKIP parties in mainland UK in 2015. In total 57% of candidates provided a full or partially completed survey, giving a sample N of 1798. In order to measure attitudes to political representation, we draw on a battery of items that were included in both the BES and the RAB on whether Parliament should contain more or fewer MPs from a range of different social backgrounds (see Table 1). 5  Figure 1 plots the percentage of candidates and voters who said there should be 'a lot more' MPs in the British parliament from the group in question. Candidates were more supportive than voters of increasing the representation of all social groups mentioned. Candidates were most likely to want 'a lot more' women, local, working class, ethnic minority and disabled MPs. Voters were most likely to want 'a lot more' local people and working-class people and women. The gap between candidates and voters were largest on support for more women, BME, LGBT and Muslim representatives. Whereas 57% of candidates said there should be a lot more women in parliament, just 24% of voters said so-a difference of 33 percentage points. And whereas 43% of candidates said there should be a lot more ethnic minorities in parliament, just 14% of voters said so-a difference of 29 percentage points.
However, just because there is a gap between voters' and candidates' preferences, it does not necessarily indicate that elites are 'out of touch', and some may regard the gap as relatively benign. Many of the groups in question are numerically underrepresented in parliament relative to their share of the British population. Candidates may be more aware of these discrepancies than the general public-and if they are keener to address them than voters this could help to generate the political will to make parliament a fairer reflection of the society it is there to represent. But these gaps could also reflect differences in a broader outlook towards which groups in society ought to be well represented. A potentially more challenging interpretation is that demands for cosmopolitan versus nativist representation might indicate clashing conceptions of the polity and the very purpose of democratic institutions and the state. From this perspective, at their extremes, cosmopolitan versus nativist demands tion of their colleagues' views or where they think the leadership of the party would like the party to be situated and so would still reflect a gap with voters even if it is insincere.
Footnote 5 (continued) for representation relate to world views that are essentially irreconcilable and not only pit some sections of society against one another, but also pit some sections of society against political elites.

Attitudes Towards Representation
To investigate these issues our analysis proceeds in two steps. First we run an exploratory factor analysis (EFA) on the voter data and candidate data separately to examine the plausibility of a two-dimensional factor solution, which distinguishes between cosmopolitan preferences for the representation of diversity in parliament and on the one hand and nativist preferences for representation on the other. Secondly, we run a latent class analysis (LCA) to identify the distinctive subgroups of the population that share common preferences for nativist representation and cosmopolitan representation. This approach allows us to examine what proportion of voters favour, for example, cosmopolitan representation, and how this compares to the proportion of candidates.
To carry out the EFA we let the survey items freely load on any latent factor in the data, be it one, two, or more and extract the factors using a principal-components factor method. A large majority of the resulting factor loadings in Table 2 conform to the expectations of our framework. Among the voters, the first factor, which we identify as the cosmopolitan dimension, primarily includes positive loadings on items that ask about the representation of disadvantaged groups such as women, people with disabilities, ethnic minorities and LGBT. Items that load on the nativist dimension, on the other hand, emphasize class and locality. This dimension also includes negative loadings on items that ask about the representation of ethnic minorities, Muslims and LGBT. The results of the EFA are therefore consistent with our first hypothesis that attitudes toward representation of different groups tap into distinct latent preferences. The pattern with respect to candidates is similar, though the nativist dimension is unsurprisingly weaker.
Interestingly the different items do not uniquely identify the different dimensions of representation. For example, support for increasing the representation of the working class positively loads on to both the cosmopolitan and the nativist dimension of representation. But whereas the working class are just one of many groups that cosmopolitans would like to see represented, class is far more exclusive for nativists. Relatedly, support for increasing the representation of Muslims positively loads on to the cosmopolitan dimension but negatively loads on the nativist dimension. This is consistent with our expectations. People who hold cosmopolitan conceptions of representation want to see an increase in the representation of diversity in parliament, across all different social groups. By contrast, people who hold a nativist conception of representation only want to see an increase in the representation of local and working class MPs, and are more ethnically exclusive in their preferences as illustrated by their opposition to representation of Muslims and ethnic minorities.
So, how many voters hold a nativist preference for representation? And how many candidates do? In order to answer these questions we carry out a latent class analysis. LCA is similar to factor analysis in that it identifies latent variables on the basis of multiple empirical indicators. However, whereas factor analysis is based on similarity (correlations) between indicators, LCA is based on the similarity between people's response patterns on the different indicators. Since nativism is conceived as a distinctive group of people who simultaneously score high on some measures (e.g. more working class MPs) and low on others (e.g. fewer Muslim MPs), LCA is the most appropriate technique for determining the extent to which this preference for representation exists in the research population. Table 3 shows the goodnessof-fit statistics for different model specifications. There is a sizeable reduction in the Bayesian Information Criteria (BIC) between a two class and three class model, after which there is relatively little improvement. We therefore opt for a three-class solution. 6 1 3 The conditional probabilities for each of the latent classes are shown in Table 4. Each class exhibits a distinctive preference for the representation of different social groups in parliament that very closely matches our theoretical expectations. The first class, which comprises 50% of the research population, corresponds closely to the expected preference for cosmopolitan representation. People in this class are very likely to say that the representation of all social groups in parliament should be increased, and thus clearly favour greater social diversity in parliament. By contrast, the second group, which comprises 20% of the research population, have more ethnically exclusive preferences which corresponds closely to the expected pattern for nativist representation. People in this class are very likely to say that the representation of local MPs and working-class MPs should be increased, but that the representation of LGBT, ethnic minorities and Muslims should be decreased. Lastly, the third group, which comprises 29% of the research population, have a more status quo preference for keeping things more or less as they are. Whereas the vast majority (80%) of candidates support cosmopolitan representation, just under half (47%) of voters do (see Fig. 2). By contrast, whereas 22% of voters support nativist representation only 5% of candidates do. There is thus a stark representation gap between the type of groups that candidates want to see in parliament and the type of groups that voters want to see. This provides preliminary support for the hypothesis that before the referendum was called political elites were out of step with the preferences of ordinary voters on issues of political identity.

Preferences for Representation: An Elite Mass Comparison
Given that there are substantial differences between the social profile of candidates and the wider population from which they are drawn, from an egocentric descriptive representation point of view one reason why there is a representation gap could just be because candidates and voters look very different from each other with respect to certain demographic characteristics. However, given that candidates are overwhelmingly white and male and able bodied, demographic differences alone would appear ill-equipped to explain why candidates are so much more likely than voters to support the increased representation of women, ethnic minorities and disabled people.   Table A1 (in Online Appendix) presents the results of an ordered probit model for voters and candidates on a selected number of the representation items, for which we have corresponding demographic data, and their support for cosmopolitan and nativist representation. 7 The average marginal effects for each of the covariates are shown in Fig. 3. There is some evidence to support the idea of egocentric descriptive representation. By and large, people of a given social characteristic are more likely to support the increased representation of that social group than people with different social characteristics. So for example, women are more likely than men to support the increased representation of women in parliament, people with a working class identity are more likely to support the increased representation of working class MPs than people with a middle class identity, and so on. The magnitude of these mimetic effects are particularly pronounced for religion, where both Muslims and Christians are particularly keen to see the increased representation of MPs from their own religious background. Though whereas Muslims are also more likely to support the increased representation of MPs from other faiths, such as Christians; this is not reciprocated and Christians are less likely to support the increased representation of Muslims. 8 Moreover, controlling for the demographic characteristics of individuals, candidates are significantly more likely than voters to support the increased representation of all social groups (with the exception of Christians). This is particularly pronounced with respect to the representation of women, ethnic minorities and Muslims. From the magnitude of the coefficients in Fig. 3, we can also see that the marginal effect of being a candidate is greater than that of being a woman on support for the increased representation of women in parliament. The marginal effect of being a candidate is also greater than that of being an ethnic minority on support for the increased representation of BAME groups. There is thus a clear divide between the preferences of candidates and voters.
But beyond this pattern of egocentric preferences for descriptive representation there are also some broader patterns of note indicating ideological preferences for descriptive representation. Women are keener than men to support the increased representation of all social groups, particularly with respect to ethnic minorities and young people. Similarly, young people are much more supportive than older people of increasing the representation of ethnic minorities and Muslims. Interestingly the magnitude of the age effect on support for more Muslims in parliament is somewhat larger than it is for the representation of more young people. By contrast, Muslims and Christians are much less likely than people with no religion to support the increased representation of women, young people and working-class MPs. These findings illustrate the limits of thinking about descriptive representation purely in egocentric terms.
Next, we therefore consider descriptive representation in more ideological terms, distinguishing between support for cosmopolitan and nativist representation. To do so we use the posterior probabilities from the LCA to assign voters and candidates to one of the latent classes. Figure 4 displays the average marginal effects from a multinomial logistic regression, with the same set of demographic controls used previously (for full results see Table A1 in Online Appendix). There are clear demographic divides between those who favour nativist representation and those who favour cosmopolitan representation. Women, young people, graduates-and particularly candidates are much more likely to hold a cosmopolitan view of representation. By contrast older men and those with few qualifications are much more likely to hold a nativist view of representation-and are much more likely than political elites, from all political parties, to do so. 9 Controlling for these factors, class and ethnicity are only weakly related to support for different types of representation. People with a working-class identity and ethnic minorities are less likely to support the status quo, but are not distinctively in favour of either nativist or cosmopolitan representation. 10 Nativist S tatus Quo Cosmopolitan Fig. 4 Support for cosmopolitan and nativist representation, AME 9 These demographic divides are much more evident among voters than they are among candidates (see Tables A3 and A4). The main differences among candidates are that women and young candidates are more likely to hold a cosmopolitan view of representation than favour the status quo, and that old candidates are more likely to hold a nativist view of representation than favour the status quo. 10 With respect to ethnicity we should treat the results with caution due to small sample sizes within specific ethnic groups. There is suggestive evidence of considerable heterogeneity amongst different ethnic groups, with black ethnic groups much less supportive of ethnically exclusive nativist representation than This disconnect between candidates and voters is evident even when we subset on party affiliation/identification. 11 However, the magnitude of the gap varies somewhat between parties. From Fig. 5 we can see that Conservative candidates are much more likely to favour cosmopolitan representation over nativist representation than Conservative party identifiers in the population at large. There is thus a big gap between the party elite and the party's supporters. This findings should come as no surprise, given how then Conservative Prime Minister David Cameron was viewed by many in the party rank and file at the time as being too socially liberal (Webb & Bale, 2014). Similarly, there is a large representation gap between UKIP candidates and UKIP party identifiers. Even though the political elite in UKIP are relatively nativist, they are still far less nativist than their supporters. This finding is consistent with the well documented difficulties that Nigel Farage faced as party leader with the expression of more extreme views by a significant number of supporters, which attracted a wave of negative publicity (Goodwin & Milazzo, 2016, pp. 46-47).
Labour candidates are also more likely to favour cosmopolitan representation over nativist representation than their base; but the difference between political elites and party supporters is not quite as pronounced as it is for the Conservatives (or UKIP).

Fig. 5 Support for cosmopolitan vs nativist representation by party affiliation, AME
Footnote 10 (continued) Asian or Chinese ethnic groups. The lack of a relationship with working-class identity is not surprising given the close relationship between social class and educational attainment and the significant role that educational attainment plays in shaping political attitudes in the UK. 11 For clarity and ease of interpretation we specify a logistic regression model and focus just on the comparison between preferences for nativist representation (1) and cosmopolitan representation (0). However, we should bear in mind that a lot of the more nativist inclined Labour identifiers may already have moved away from the party towards UKIP (Evans & Mellon, 2015) or abstention (Heath, 2018) by 2015 and so this disconnect perhaps underplays the tensions that had previously existed between the political elites and their traditional supporters.
Although there is some variation in terms of the magnitude of the representation gap, there is a significant gap between elites and party supporters among all political parties apart from the Greens and the Nationalists (SNP and Plaid Cymru). Thus, there is clear evidence that the nativist and cosmopolitan divides that have come to public attention in the UK in the aftermath of the EU referendum were present well before it was even called. These divides cut across different social groups-separating the young and the well-educated from the old and those with few qualifications; but also separating voters from political elites, even within party lines. Whereas among candidates, there was almost universal agreement that parliament should reflect greater diversity in all walks of life; among voters this view was much more hotly contested. Indeed, the preference for nativist representation among a small but significant group of voters can in some sense be seen in conflict with the changes that have recently taken place in the UK parliament, which has witnessed a pronounced increase in the number of women and ethnic minority MPs but a decrease in the number of white working class male MPs. Although we cannot tell whether these changes in the composition of parliament have softened nativist attitudes among voters or hardened them, we can see that before the EU referendum there were a non-trivial number of people who were resistant to the dominant direction of travel that was almost universally supported by political elites, which emphasized the representation of ever greater diversity in parliament. This pool of nativist voters were potentially fertile territory for populist mobilization.

Populist Sentiment and Brexit
To investigate this possibility our attention now turns to the attitudinal and behavioral correlates of these dimensions of representation. First, we consider the relationship between support for nativist (and cosmopolitan) representation and support for anti-elite attitudes. Then we consider the relationship between support for nativist (and cosmopolitan) representation and support for populist movements, such as the vote to leave the European Union. To measure anti-elite attitudes we use the following question: "Politicians don't care what people like me think." Our expectation is that people who hold a preference for nativist representation will be more likely to agree with this statement than people who favour-and are thus happy with-the status quo. By contrast, citizens with a preference for cosmopolitan representation may perceive a gap between the populace and the political elite running in the opposite direction. From a cosmopolitan perspective the contemporary British political elite is somewhat retrogressive in terms of group representation and is over-representative of white middle-class men. Thus, despite their differences we anticipate that voters with a preference for both nativist and cosmopolitan representation will feel more social distance from MPs and express more anti-elite sentiment than those 1 3 who prefer the status quo. To test these expectations we specify an ordered probit with a number of controls. Figure 6 plots the average marginal effects for each of the covariates (see Table A2 in Online Appendix for full results). Anti-elite attitudes are more common among men, working class identifiers, ethnic minorities, those who hold left wing and socially authoritarian views and those people who either do not identify with any political party or identify with a fringe party like UKIP or the Greens. However, even after controlling for all these factors, people who support nativist representation are more likely to think that politicians don't care about them than people who support the status quo. The magnitude of the effect is substantively large. We also find that those who prefer cosmopolitan representation are more likely to express anti-elite attitudes than those who favour the status quo. Thus, a preference for both nativist and cosmopolitan representation is an expression of dissatisfaction with a gap between the current make-up of the legislature and the desired configuration that that generates a disconnect between citizens and their representatives.
These results indicate that people who support nativist representation are not only much more likely to do so than political elites; but are also more likely to think that political elites do not care about them. Part of the reason for this may be that they regard politicians as part of the cosmopolitan elite, who care little for the groups of voters that nativists want to see represented. But part may also be that nativism embodies a sense of victimhood, where nothing is as good as it used to be or should be, and so nativists are inclined to have a somewhat negative view of the political (and social) world in general, regardless of the reality. Equally populism and nativism are common bedfellows as "populism needs a focus for its animosity" (Iakhnis This implies that those who prefer nativist representation would appear prime candidates for populist mobilization. Interestingly citizens with cosmpolitian preferences for political representation express a similar, if slightly weaker, level of antielite sentiment to those with nativist preferences. Cosmpolitans also perceive a gap between who our representatives are and who they believe should be better represented, but their preference is for greater diversity of representation across a range of under represented groups. If politicians and parties were to adopt more nativist strategies for political recruitment this may provoke a backlash from those who prefer cosmpolitan representation. However, our analysis shows that the representation preferences of political elites are closely aligned to voters who prefer cosmpolitan representation and thus the impact of the representation gap on populist sentiment is likely to be expressed politically from a nativst rather than cosmopolitan perspective. To this end we next consider support for Brexit. We specify a logistic regression model, where the dependent variable is support for Brexit (vs Remain). Figure 7 plots the average marginal effects for each of the covariates (see Table A2 in Online Appendix for full results). Brexit has widely been described as a populist revolt-yet relatively little attention has been given to the political sources of grievance that may have played a role (though see Iakhnis et al., 2018). Even though at the time of the survey the referendum was a distant and uncertain proposition, we can see that the age and educational divides between 'Leavers' and 'Remainers' were already clearly evident. Even after controlling for a host of social and political factors, we also find that people who favour nativist representation are significantly more likely to support Brexit than those who favour the status quo, and that people who favour cosmopolitan representation are significantly less likely to support Brexit than those who favour the status quo.
These results indicate that identity concerns about which groups of people in society should be represented politically have wider political resonance in a political context where party identiy has diminished whilst new identities associated with attitudes to globalisation coalesced around Remain/Leave support (Heath & Richards, 2019;Hobolt & Tilley, 2019). The contrasting identities of cosmopolitan liberals versus those with a more inward looking emphasis on Englishness described by Jennings and Stoker in their tale of two Englands are reflected in demands for political representation and are predictors of hostility towards the EU (Jennings & Stoker, 2016). In common with other analysis of support for Brexit we find that older, working class voters with few qualifications are the most likely supporters (Hobolt, 2016) but we find that in addition to these factors holding nativist demands for political representation (preferring the greater representation of local and working class MPs and few ethnic minoritites and muslims) are also associated with Brexit support.
These effects still hold even when we control for attitudes towards immigration (see Figure A3 in Online Appendix which illustrates the average marginal effects for selected variables). 12 People who think immigration should be reduced are much more likely to support Brexit. But even taking this into account, people who hold a nativist preference for representation are still more likely to support Brexit. This indicates that our measure of nativism moves beyond just anti-immigrant sentiment and taps in the question of who voters want to see as representatives of the people.

Conclusion
The findings in this article have important implications for understanding the ways in which voters and politicians think about representation. Using the UK as a case study we have found that attitudes to descriptive representation are motivated by both egocentric and ideological concerns. In addition to preferring representatives who look like themselves in various respects, voters also prefer legislatures to descriptively represent groups that fit with their political attitudes more broadly. These preferences reflect different world views of which groups should be well represented. People who support cosmopolitan representation are positively inclined to support the increased representation of all under-represented groups, but they give priority to increasing the representation of women, ethnic minorities, the young, Muslims, LGBT and people with disabilities. By contrast people who prefer nativist representation prioritize increasing the number of local (white) working class MPs and want to see fewer Muslims and ethnic minorities.
Although the UK political elite is generally supportive of cosmopolitan representation (80% of candidates), voters are much less likely to do so (47% of voters). By contrast voters are much more likely to support nativist representation than candidates (20% vs 5%). There is thus a stark gap between elites and voters, and elites are very out of touch with the more nativist elements of society. Furthermore, we find that this group of voters who support nativist representation are more likely to think that politicians don't care about what people like themselves think and were more likely to support Brexit. Supporters of cosmopolitan representation were also more likely than those who prefer the status quo in parliament to think that politicians don't care what people like themselves think, reflecting the representation gap in terms of their preferences. Thus, a nativist political solution to the representation gap would likely trigger a backlash from the larger section of the electorate that supports cosmopolitan representation. Those who preferred cosmopolitan representation were less likely to support Brexit than supporters of the status quo in parliament reflecting an internationalist perspective.
We cannot assess whether the divergence between the preferences for group representation among the political elite and the mass public play a role in fueling populist anti-elite sentiment but there is certainly a lacuna that is ideally suited for populist parties to fill. This lacuna has emerged as the established social democratic political parties in many countries have moved from mass membership parties rooted in local working-class communities to professional catch all parties designed to capture a greater share of the growing middle-class vote. This shift in priorities has not only impacted the policy programmes of leftwing parties; but also their personnel who are increasingly drawn from middle class backgrounds (Heath, 2015).
A great deal of research has documented the policy failures of mainstream parties to meet the needs of the 'left behind' communities of post-industrial societies but considerably less attention has been paid to their descriptive representation and how this may have fueled growing populist sentiments. The rise of nativist populism is a serious challenge facing contemporary liberal democracies; populism is by definition a simplistic approach to addressing complex problems and when combined with nativism threatens the fundamental tenets of liberal democracy, especially the protection of minorities. We know that sectors of the electorate considered left behind by globalization are those most likely to be supporters of the populist radical right and now we know that this cosmopolitan/nativist divide extends to attitudes towards political representation. The decline of working-class representation as parties of the left targeted the votes of the growing middle classes has perhaps provided the environment where nativism has become coupled to demands for working class representation for a significant minority of the electorate. The failure of the major parties to select proportionate numbers of working-class candidates provides fertile ground for populist nativists to set a white working-class identity against a diversity of representation. Our analysis of the British case shows that 'who' our representatives are might be an equally important issue underlying the divided politics of many contemporary liberal democracies as the 'what' they do when in office. This data was collected in 2015 just prior to the 2016 EU referendum vote, which was predicted to be a secure win for the remain side but instead resulted in the UK's exit from the EU radically transforming the political landscape both in its international relations and internal party politics. After a long period from the early 1990s of relatively centrist politics in the UK populist politics, on the right and left, gained massive ground and our analysis suggests that dissatisfaction with group representation was a significant feature of the discontent expressed by a sizeable minority of the electorate. Such seismic political shocks have subsequently become common place in Western democracies and alienation from formal political representation may well be part of the explanation. Our analysis suggests that to reinvigorated liberal democracy attempts to strengthen and legitimize institutions should be accompanied by serious efforts to mobilize working class communities into elected office.