The victims of neoliberal globalisation and the rise of the populist vote: a comparative analysis of three recent electoral decisions

The victims of neoliberal globalisation and the rise of the populist vote: a comparative analysis... Abstract Recent presidential elections in the US and Austria as well as the referendum on Brexit in the UK delivered victories or near-victories for populist right-wing candidates or agendas. In all three cases, globalisation and European integration were blamed for higher immigration and pressure on public services, deindustrialisation and job losses, and the attack on traditional values by cosmopolitan elites supported by traditional centre parties that have been unable or unwilling to control those processes. While election analysts seek to explain voting behaviour with socio-demographic characteristics of individuals, individual voting preferences also depend on the geographical context in which decisions are made. This article thus examines how long-term, regional structural economic changes, the varying impact of the Great Recession on the rise of and recovery from regional unemployment and current regional economic conditions, such as unemployment and welfare benefit losses, affect regional vote shares. In addition to those economic conditions, we examine the impact of immigration and urban size on populist vote shares. We show that regions with low, but rising immigrant shares, old industrial regions, smaller regions, those whose labour markets were exposed more and recovered less from the Great Recession, and those with high unemployment rates and benefit losses exhibited higher populist vote shares. These results are largely consistent across the three case study countries. Introduction While populist parties did not suddenly burst on to the political scene after the Great Recession in the 2010s, a number of recent electoral decisions in the US and Europe returned wins or near-wins for right-wing populist candidates and their agendas. Many of the recently successful right-wing parties with a strong emphasis on ‘nativism’ and populism were either founded or moved from a liberal economic focus to a nativist focus in their policies in the 1980s or early 1990s (Mudde, 2007). While economic policies do not make up the core of populist party programmes, the rise of the populist vote nevertheless coincides with the rise of neoliberal globalisation and its impact on employment and wage structures in the countries of the Global North, coupled with an abandonment of the traditional working class and working class areas by traditionally left-wing parties (Eribon, 2009). This representative void may explain the shift in the communist vote in France or the Socialist vote in Austria to the populist right for instance. In addition to those gradual shifts towards the populist vote, we also witnessed a recent surge in the level of that vote relative to previous elections, which may be related, in part, to recent economic hardship produced by the Great Recession. A large number of competing and complementary conceptual frameworks have been employed to explain the rise of populist parties offering macro-, meso- and micro-level explanations of the demand for and supply of populist agendas, as well as the constitutional arrangements governing the electoral process (Golder, 2016; Mudde, 2007). These studies tend to focus on cross-country comparisons using individual survey data. However, while cross-country studies can highlight the role of broad economic, institutional and historical differences that may influence voting behaviour, and studies focusing on micro survey data can tell us something about the stated preferences of individuals, studies at the national scale do not account for pronounced geographical differences in economic, social and cultural contexts that generate demand for the populist vote (Golder, 2016). On the other hand, analysis of individual survey data tends to assume preferences and psychological profiles of respondents to be exogenous to the context in which those preferences are stated. Recent electoral outcomes involving populist parties or agendas were characterised by substantial geographical differences in the shares of the populist vote. By studying the geographical pattern of three electoral decisions, the presidential elections in the US and Austria as well as the Brexit vote in the UK, we want to explore whether geographic differences in selected demand factors for populist agendas offer reasonable explanations of electoral outcomes in those three countries. Focusing on the subnational rather than national scale of analysis should produce a closer matching of the actual opportunities and constraints that individuals face as a result of neoliberal globalisation and increasing economic and political integration, as well as the geographically varying impact of the Great Recession. While we do not claim to provide a comprehensive explanation of the rise of populist parties and candidates in Europe or the US, we offer a first-cut analysis of the role of regional economic differences to explain the populist vote shares, and through that flagging the potential importance of subnational electoral studies to political scientists. While they recognise heterogeneity among the preferences of individual voters, they ignore that those preferences are driven by growing up in families whose political preferences are shaped, in part, by socio-economic opportunities and constraints, coupled with the formation of cultural practices offered by the places in which they live. In our analysis, we thus focus on demand-side explanations of the populist vote, specifically factors that reflect the economic conditions of regions within the three countries we study. While most studies focus on unemployment rates, income or occupation to operationalise economic conditions, we distinguish between structural and cyclical changes as well as current economic conditions. Long-term structural decline as a result of increasing global competition and integration should result in a gradual increase in demand for populist agendas in those regions. We consider old industrial and mining regions to fall into that category. But voting decisions will also be influenced by cyclical factors, in particular the differential impact of the Great Recession on employment opportunities. And finally, voting should be affected by the current economic situation that individuals face. In addition, because all three electoral campaigns were fought on the back of ‘immigration crises’ and the lack of national control of it, we also examine the importance of migration to explain the share of the populist vote. More specifically, we ask the following questions: First, despite different national economic contexts, does the negative impact of neoliberal globalisation and tighter economic integration on old industrial areas result in higher populist vote shares? Second, to what extent does the impact of the Great Recession on regional employment affect the populist vote share? Third, how do the current regional economic conditions in terms of unemployment and welfare losses affect the populist vote share? Fourth, do regional differences in immigration rates influence the populist vote share? As our study is based on the cross-sectional data, we can explore how well our economic variables explain regional voting patterns, but are limited in establishing clear causal links between economic macro patterns and individual decision making. The article is structured as follows. The second section discusses the theoretical arguments for the rise of populism and their link to economic geography, while the third offers a brief overview of the gradual successes of populist parties and candidates in the three countries. Section four explains how the theoretical arguments are implemented empirically through a discussion of methods and data, the fifth section presents the main findings and section six concludes the paper. Populism and economic geography ‘Neoliberal globalisation’ and ‘populism’ are overused and conceptually stretched terms, but we use them to organise our empirical analysis and potential explanations of recent electoral decisions. Under ‘neoliberal globalisation’ we understand a specific form of globalisation in which the benefits of market integration become unquestioned truth (Harvey, 2005; Mouffe, 2016; Müller, 2016; Sheppard, 2015). In reality, the benefits and costs of globalisation are distributed unequally between social groups and different territories and so open up a cleavage between those in favour of more and those in favour of less integration (Kriesi et al., 2006). Political elites do not question seriously the perceived positive effects, thus those bearing the cost feel increasingly misunderstood, unheard and unrepresented by those political elites. While neoliberal globalisation does not cause the rise of populism, the undifferentiated belief in the benefits of globalisation by traditional parties on the right and left of the political spectrum and lack of political representation of those harmed by the process generates a representational void that is increasingly filled by populist parties, even if economic policy is not generally part of their core programmes. To understand why globalisation may result in the rise of populism we need a working definition of ‘populism’. Populist right-wing parties, candidates and their agendas1 have three characteristics. They are (i) anti-establishment, (ii) nativist, and often (iii) authoritarian. The populist right-wing vote is directed against perceived corrupt elites who are supposedly not representing the will of ‘the people’, understood by populists as a homogenous group of nationals with the exclusive rights to national resources (Müller, 2016). Because of the focus on ‘the pure people’ and the exclusive rights to resources attributed to them, populists are also anti-pluralist. They reject the need to compromise in order to reconcile differing group interests (Mudde and Kaltwasser, 2013). In those respects, a number of recent elections in the UK, US, Austria, the Netherlands, France, Hungary or Poland can therefore be interpreted as mandates for populist right-wing candidates promising to keep others out, taking control over their own resources and getting one over on the elites that supposedly do not know what the real people want and need. The question is what factors and processes explain the rise of right-wing populism?2 Here we briefly review some of the most important theories and discuss how they may help us understand geographic variation in the populist vote. Broadly, explanations can be separated into demand for populist parties, the supply of populist political programmes, messages and parties and the institutional context in which those parties operate (Mudde, 2007). The focus in the empirical literature on populism is on demand-based explanations that can broadly be grouped into those emphasising the role of economic and those emphasising the role of cultural changes as main cause for the rise of demand for populist agendas (Inglehart and Norris, 2016). One influential argument is based on the idea that recent developments and processes such as globalisation, deindustrialisation, rising inequality or the transition to post-Fordism generated a new group of Modernisierungsverlierer (losers of modernisation). While echoing Lipset’s (1960) analysis of the 1930s, where rapid industrialisation generated fears of downward social and economic mobility among the petit-bourgeoisie, small entrepreneurs, shopkeepers, merchants, self-employed artisans and independent farmers squeezed between big business and organised labour, that made them susceptible to fascist and extremist ideas, recent changes generated a new underclass that forms a new electoral base of populist right-wing parties (Betz, 1994; Esping-Anderson, 1990). While continuing to appeal to parts of the petite-bourgeoisie, populist right-wing parties now find new demand among the low-skilled, poorly educated, blue-collar working class whose jobs have been eradicated by globalisation, automatisation, deindustrialisation, have become more contested through competition from low-skilled immigration or devalued through stagnating or declining wages exacerbated by blue-collar union decline, austerity policies and an increasing inability of nation states to regulate big business or immigration (Betz, 1994; Inglehart and Norris, 2016; Piketty, 2014). These economic changes are translated into a new cleavage structure (Kriesi et al., 1995). In the past, European party systems have been dominated by economic (class) cleavage. Globalisation and deindustrialisation have led to a decline in class voting and partisan identification, increased political alienation among certain segments of the population and reduced trust in the political elite (Golder, 2016, 488). Changes in the occupational structure and resulting social group affiliation, as well as increasing secularisation on the centre-right, eroded the traditional voter pool, which meant that traditional centre and left parties searched for voters elsewhere and abandoned, to a large extent, their traditional base (Eribon, 2009). This process increased electoral volatility and created a pool of floating voters rather than traditional partisan loyalties generating a demand for representation that populist parties were able to capture (Betz, 1994; Minkenberg, 2000). Harnessing support from an economically diverse group of small business owners, white-collar workers, skilled and unskilled blue-collar workers, as well as the unemployed, necessitated identification through cultural, not economic values (Evans, 2005). The cultural cleavage that is forged rests on group identities based on nationality and ethnicity. Ethnic minorities and immigrant groups are blamed for deteriorating conditions, loss of jobs and increasing pressure on welfare services. While the literature on Modernisie rungsverlierer focuses on the long-term impact of economic changes on voting behaviour, a second set of work focuses on the role of crisis (Kriesi et al., 1995; Taggart, 2000). During economic crises the pressure on and competition for jobs and welfare services increases and hence, discussions about how those resources should be allocated among different groups become even more important. Immigrants and ethnic minorities are easier to blame than complex relationships between technological change, globalisation, neoliberal austerity policies and resource constraints.3 If we were to follow the logic of this argument, we would expect increasing vote shares for populist candidates during economic crisis and in regions affected most negatively by a crisis. However, the empirical evidence is mixed. While Lubbers et al. (2002) find a positive relationship between unemployment and voting for populist parties at the individual level, they fail to identify the link at the macro level. Theoretically, it is possible to argue that during times of crisis, voters fall back on traditional parties, as they are trusted more on economic issues than populist parties who mobilise their voters along cultural commonalities (Ivarsflaten, 2005). Golder (2003) established a more direct link between crisis, unemployment and immigration, and finds that unemployment is positively related to populist vote shares in countries with high immigration but less so if immigration is low. The idea of cultural cleavage as reaction to socio-economic changes induced by economic restructuring processes since the 1970s is picked up and re-formulated more broadly as ‘cultural backlash thesis’ by Inglehart and Norris (2016). They interpret the rise of the populist vote as cultural counter-revolution against the rise of progressive values, including increased tolerance towards sexuality, same-sex marriage, LGBT rights, ethical norms, open-mindedness towards migrants, refugees, foreigners, multicultural lifestyles, foods and travel, cosmopolitan support for international cooperation, humanitarian assistance, and multilateral agencies like the UN or EU, environmental concerns, race and gender equality and human rights (Inglehart, 1990, 1997; Norris and Inglehart, 2009). The transition from materialist to post-materialist values is accomplished through generational replacement and is linked to the rise of Green parties as well as progressive social movements and transnational activist organisations. As the share of post-materialists increases in the population, they brought new issues into politics, resulting in a declining emphasis on social class and redistribution, and an increased polarisation around cultural issues and social identities. Those favourable towards progressive social change and humanistic values tend to be economically secure and better educated. The rise of those values in the 1960s and 1970s was opposed early on by older traditionalists who felt threatened by the erosion of values. In the context of Western European societies, it is relevant that younger generations tend to grow up with greater economic security, and people growing up in secure circumstances have been shown to be more open-minded, socially tolerant and trusting, more secular and accepting of diversity. In addition, increasing participation of women in the workplace and higher education altered traditional gender roles. Traditional values were therefore held increasingly by the old, men and uneducated. In this view, xenophobia and hostility and intolerance towards migrants, ethnic and racial minorities are only part of a wider cultural backlash against social and cultural change. Empirically ‘culture’ is measured as educational attainment, trust in governance, anti-immigrant sentiments, authoritarian values or right-wing ideology (Inglehart and Norris, 2016). But scores on those variables are not independent of the economic and social positions of the respondents or the positions of families and/or regions and neighbourhoods they grew up in. A wider problem is that empirical results produce contradictory evidence for both the economic and cultural theses. In part, this can be explained by the difficulty in isolating cultural from economic effects, but it may also result from cross-sectional analysis of national election results. Considering substantial regional variation in electoral outcomes (Becker et al., 2016; Golder, 2016), focusing on whole countries appears problematic. For instance, while we may obtain a positive correlation between the share of immigrants and populist party vote at the national scale this does not have to be the case at the subnational scale, if immigrants are concentrated in large metropolitan areas that, overall, benefit from globalisation and economic integration, while rural areas (with very few immigrants) suffer. In this case, the relationship between immigration and populism may turn negative at the subnational scale. Analysing three electoral outcomes at the subnational scale should generate a tighter fit between the theoretical arguments linking economic, social and cultural contexts and voting outcomes. There is now some literature on regional differences in the Brexit vote (Becker et al., 2016; Goodwin and Heath, 2016; Harris and Charlton, 2016; Hobolt, 2016; Los et al., 2017), but this work tends to focus on socio-demographic variables, psychological variables, immigration, survey data or current economic conditions and neglects the role of long-term structural decline or the impact of and recovery from the Great Recession as potential explanatory variables for regional Brexit vote shares. For the US, Autor et al. (2016) examine the potential role of import competition from China on the increase in the Trump vote share, but they do not further examine long-term structural decline and cyclical effects as potential causes of higher Trump vote shares. We do not know of any studies that attempt to explain regional variation in the outcome of the Austrian presidential elections yet. Rather than restricting our analysis to one country, the comparison between three elections allows us to examine whether subnational economic changes affect the electoral outcomes in similar ways, despite differences in national political landscapes, differential exposure to and impacts of modernisation processes on the economies, sectors and occupational structures of those countries. While the Brexit vote and the Austrian presidential elections were decided by vote shares of the national electorate, the US presidential election depends on winning the electoral college of individual states. In this sense, supply-side factors, that is party strategies to shift non-Republican votes, need to target particular groups of voters in states that can be turned by them. We explore this by conducting separate analysis for individual regions within the US. Examining changes in populist vote shares allows us to examine whether regional differences in the long-, medium- and short-term economic conditions have a stronger impact on the populist electoral base or their floating voter shares (Arzheimer, 2011). We engage with and add to this literature in the following way: First, we add three empirical studies at the subnational scale to the sparse empirical literature on regional differences in populist vote shares. This is important because the large geographical variation in populist vote shares will cause problems for national-scale analysis if the postulated roots of increasing populism such as globalisation or deindustrialisation affect regions and people living on those regions differently. Second, attempting to explore regional voting patterns in three countries through identical model specifications allows us to examine whether regional context has similar effects on voting patterns, despite national historical and institutional differences. Third, in the empirical analysis below we favour economic over cultural explanations of rising populism. Although we readily admit that the rise of populism cannot be reduced to economics and will always have a social and cultural dimension, we follow Eribon (2009) and believe that the structural position of individuals in a socio-economic system will determine, in part, their social group membership and associated group-specific cultural practices, so that the influence of culture and social group membership cannot be independent of and isolated from economic context in which individuals grow up. Also, because our empirical work is based on cross-sectional data and ‘cultural’ and ‘economic’ variables correlate highly at the regional scale, we have little choice but to focus on one of the two. Fourth, we separate empirically the impact of modernisation and economic crisis on rising populist voter potential by distinguishing between long-term economic changes, the effects of the Great Recession and current economic conditions in a region. Trends of populist right voting in the three countries To set the scene, we provide some descriptive historical and geographical information on the populist vote shares in those countries. In the UK, Brexit was strongly pushed by the UK Independence Party (UKIP), in the US, the Republican Party and Tea Party moved increasingly towards the right and in Austria, the Freedom Party (FPÖ) enjoyed considerable vote gains since the election of Jörg Haider as their party chairman in 1986. Because of the first past the post system of the UK electoral system, UKIP only managed one seat in the House of Commons in the 2015 general election. However, this masked the rise of UKIP as an electoral force in the UK. Founded in 1993, UKIP increased its percent of overall votes in general elections from 0.3 in 1997 to 12.6 in 2015. At the same time, it rose to become the strongest British faction in the European parliament, where they obtained 27.5% of the vote and 24 of the 73 seats to be won in the 2014 EU parliamentary election. They were able to gain votes from an increasingly disenfranchised (and socially conservative and inward looking) working class neglected by New Labour and social conservatives on the right. The accumulated pile of populist votes was set on fire by increasing concerns over immigration from the EU accession countries ignited by Tony Blair’s decision not to follow other EU countries and pose restrictions on immigration from Eastern European EU countries, and kindled by David Cameron’s conservative party by promising to reduce net migration to ‘tens of thousands’ prior to the 2010 elections (Evans and Mellon, 2016; Ford and Goodwin, 2017). While a large share of UKIP voters comes from a conservative base, Brexit campaigning by UKIP tended to concentrate on the old industrial, labour voting areas and the workers left behind by globalisation and integration. Coupled with a lacklustre commitment of Labour Party leader Jeremy Corbyn to Europe, old industrial areas then voted overwhelmingly for Brexit (Goodwin and Heath, 2016; Johnson, 2015; Shafique, 2016). Although there are some question marks whether Donald Trump can really be seen as populist (Oliver and Rahn, 2016) or not (Singh, 2017), roll call vote studies show increasing polarisation and shift towards the right of the US Congress and Senate since the late 1970s (McCarty et al., 2016; Pool and Rosenthal, 2009). In that sense, Trump built on a right-moving Republican base and harnessed growing sentiments against the effects of neoliberal globalisation, high debt, low employment and wages and rising inequality by blaming the uncaring and unknowing elites, as well as easy scapegoats (Mexican immigrants and Chinese exports) packaged into simple slogans that pitch the true Americans against the rest of the world (Blyth, 2016; Blyth and Matthijs, 2017). During the election, the usual Republican voters turned up, but Trump did especially poorly with women and those with college education, did better than previous Republican candidate Mitt Romney with minorities, and did exceptionally well with white voters without a college degree, in particular in the old industrial heartland of the Midwest and, to a much lesser extent, among the manufacturing workers in the South (relative to Romney’s vote shares) (Tyson and Maniam, 2016). In the end, Trump managed to deliver 304 electoral votes to Clinton’s 227 despite losing the popular vote by approximately 3 million votes (46.1% compared to Clinton’s 48.2%). It was the capture of a large number of states in the old rustbelt of the US that included the turning of the blue states of Michigan, Pennsylvania and Wisconsin and the swing states of Iowa and Ohio that delivered the presidency. In Austria, although the President is only the head of state with de facto limited powers, the presidential election was seen as test case for the upcoming general elections in October 2017. As in the UK and the US, the first round of the election highlighted voter dissatisfaction with the status quo represented by the two centrist parties, the SPÖ (Social Democratic Party of Austria) and ÖVP (the conservative Austrian People’s Party), whose candidates obtained 11.3% (SPÖ) and 11.1% (ÖVP) of the vote, respectively. Norbert Hofer, the candidate of the populist right, the Austrian Freedom Party (FPÖ), won the first round with 35.1%, followed by the independent candidate,4 Alexander Van der Bellen, with 21.3%. Similar to the French case, voters from all parties rallied to avoid the populist right-wing candidate from winning in the run-off elections. Even so, he still managed to obtain 49.65% of the vote. Because of voting irregularities, the result of the first run-off election was annulled and in the re-run, Hofer obtained 46.21%. Hofer was able to build on the success of the FPÖ in national and regional politics since the party was transformed by Jörg Haider in 1986 into a right-wing populist party opposing the elites represented by the two centrist parties that dominated post-WWII Austrian politics, ditching former neoliberal economic and German national roots from the official party programme and replacing it with anti-immigrant, anti-EU, anti-globalisation, anti-consumerist positions to defend ‘true’, national Austrian values based on Christian heritage, law and order and welfare benefits (welfare chauvinism) from Islamic fundamentalism, immigrants and aggressive capitalism (Fallend, 2013; Müller, 2002). As a result, the FPÖ was able to increase its vote share in national elections from 5.0% in 1983 to 26.9% in 1999 and was able to win the provincial election in the Southern state of Carinthia, where it replaced the SPÖ as leading party.5 It is important to notice that this transformation and exploitation of the cultural cleavage between traditionalist and cosmopolitans was made possible by rapidly deteriorating economic conditions in the early 1980s, exemplified by the jump in unemployment rates from 1 to 4% in 1982, reflecting job losses in the nationalised steel and metal industries. This step-change in unemployment demonstrated the inability of centrist parties to deliver full employment and maintain a high-level welfare state in the face of neoliberal globalisation. While the exploitation of cultural cleavages emerges as key element in all three electoral decisions, those cleavages opened up as a result of deteriorating economic conditions for traditional blue-collar workers that now supplement the original voter potential of the right that consisted of nationalists and the petite-bourgeoisie. But in addition to those gradual shifts, the recent electoral outcomes represented a further dramatic shift towards the populist right (especially in the UK and Austria) that dovetail with further deteriorating economic conditions and austerity policies brought on by the Great Recession. The literature on the rise of populism is rather silent on those possible impacts of short-term economic changes on the populist vote share. In addition, while explanations based on traditional–cosmopolitan cleavages should be reflected in simple rural–urban differences in populist vote shares, the actual geographies of the electoral outcomes are more complex than that. Neoliberal globalisation and the ensuing deindustrialisation and immigration have had geographically varying impacts that need to be considered in macro studies of electoral outcomes to examine the contribution of local and regional context that may be more important than the national context to influence voting decisions. The neglect of the subnational scale in explaining recent electoral outcomes is even more surprising if we consider the enormous geographic variation in the populist vote share observed in those three electoral decisions (see Figure 1a–c). The populist vote shares ranged from 13 to 85% in Austrian municipalities, from 21.3 to 75.6% in British Local Authorities and from 4.1 to 95.3% in US counties. In the following, we discuss the data and methods we used to explain those differences. Figure 1. View large Download slide View large Download slide View large Download slide (a) The geographical distribution of the share of the vote for Norbert Hofer in the second run-off to the Austrian presidential elections, 2016 (municipalities). (b) The geographical distribution of the share of votes for Donald Trump in the 2016 US presidential elections (US counties). (c) The geographical distribution of the Brexit vote share, 2016 (local authorities). Figure 1. View large Download slide View large Download slide View large Download slide (a) The geographical distribution of the share of the vote for Norbert Hofer in the second run-off to the Austrian presidential elections, 2016 (municipalities). (b) The geographical distribution of the share of votes for Donald Trump in the 2016 US presidential elections (US counties). (c) The geographical distribution of the Brexit vote share, 2016 (local authorities). Data and method Data for the analysis are obtained from a variety of sources. For Austria, data on socio-demographic factors come from the Austrian Census and were obtained from Statistik Austria (www.statistik-austria.at), while data on elections were taken from a recently established database on Austrian election results (www.wahldatenbank.at). For the UK, the main data source was the Office of National Statistics (www.ons.gov.uk), while the 1980 manufacturing data were retrieved from the UK data service (http://casweb.ukdataservice.ac.uk/) and election data came from the electoral commission (www.electoralcommission.org.uk). For the US, socio-economic data were compiled from the 5-year pooled American Community Surveys, 1980 manufacturing data from the County Business Patterns from the US Census Bureau (www.census.gov), unemployment figures from the Bureau of Labour Statistics (www.bls.gov), the 2016 election results from (townhall.com/election/2016/president) and the 2012 election results from ‘The Guardian’ (www.theguardian.com/news/datablog/2012/nov/07/us-2012-election-county-results-download#data). In all three cases, local authority, county and municipality border changes had to be taken into account and were processed to guarantee that data were available for consistent geographies over time. Our dependent variable was the share of the Brexit vote in 348 local authorities in England and Wales, the share of the vote for the presidential candidate of the Austrian Freedom Party in 2122 municipalities and the share of the Trump vote in 3108 counties of the continental US. To account for the spatial differences in the populist vote, we develop the following, exploratory models:  Vote%=α0+β1IMM+β2STRUCT+β3CYCL+β4ECON+β5X+β6D+ε (1) where Vote% is the share of the populist vote, α0 is a constant, IMM represents as set of variables capturing immigration, STRUCT is a set of variables representing long-term structural economic conditions, CYCL are variables that account for the impact of cyclical economic conditions, the impact of the Great Recession on unemployment and recovery from it, ECON is a vector referring to current economic conditions, X represents socio-demographic controls such as age and gender, and population size D is a set of regional dummy variables controlling for broad regional differences in the populist vote and ε is an error term that may be heteroskedastic of unknown form. While the focus of our models is on economic context, political discourse prior to the three electoral decisions was strongly influenced by various forms of anti-immigrant statements. At the national-scale immigrants are easy scapegoats for failed economic policies, but at the subnational scale, it will depend on whether immigrants settle in regions that benefit or suffer from global economic integration and on the pace of the increase of migrant shares. Another question is whether the origin of immigrants plays a role in explaining the populist vote shares. While UKIP was concerned primarily with the perceived negative effects on public services because of migration from Eastern European EU accession countries (which was also interpreted as EU imposed loss of control of national borders), Austria appeared more worried about a perceived ‘refugee crisis’, while in the US, immigration from Latin America, in particular Mexico, became a key issue in the elections. Hence, we use different variables to represent the impact of immigration, IMM. For the UK, rather than using the share of foreign-born people, we distinguish between the share of people from EU15 countries (joined the EU prior to 2003), eight Eastern European EU accession countries, and the rate of increase in the share of residents from EU accession countries. In the US, we consider the share of the non-white Hispanic population and the increase in the share of the Hispanic population. Although most people of Hispanic origin are not immigrants, they form part of the visible minority that was singled out by Trump, so that we considered the share of this group as more important than the share of foreigners to explain the Trump vote shares. Empirically, the percent of foreign-born and Hispanics are highly correlated. In Austria, because of the small absolute numbers of visible ethnic minorities and refugees, we had little choice but choose the share of foreign-born and the rate of increase of foreign-born as our independent variables representing immigration. We are agnostic about the direction of the influence of the share of immigration variables because it is unclear what the actual effect of increased contact with immigrants is on the perception of them in the voting population. On the one hand, the contact hypothesis states that increasing contact with foreigners reduces the fear and misperceptions about them and, hence, would make it less likely to vote for an anti-immigration policy, but on the other hand, a higher share of foreigners may be considered as competition for jobs and pressure on public services, which should increase the populist right-wing vote share. Because perceived and actual pressure on public services will also depend on the increase rather than level of immigration, we include the rate of increase in immigration as well. For the UK, we look at the increase in immigrants from eight Eastern European EU accession countries, in the US, the increase in the share of people of Hispanic origin and in Austria, simply the rate of change of the foreign-born population. We expect that the rate of increase in immigration is positively related to the populist right-wing vote share. Our set of economic variables corresponds to the long-term economic changes identified as important by the ‘modernisation theorists’, the cyclical effects identified by the ‘economic crisis’ theorists, and current economic conditions that, at the regional level, are the outcome of those long and medium to short-term changes. Our structural economic variables, STRUCT, include the share of the manufacturing sectors in 1980 in the US, 1981 in the UK and 1991 in Austria,6 as well as the rate of change in manufacturing employment between the base year and the most recent year data were available. Choosing manufacturing shares from the 1980s rather than current manufacturing shares allows us to identify old industrial regions. The change in manufacturing employment was included to examine whether the pace of industrial restricting has an influence on voting behaviour. Following Eribon (2009) and Kriesi et al. (2006), who propose that a shrinking traditional working class and its political abandonment by centrist and socialist parties result in resentment and shift of traditional working class votes to the right, we would expect a positive relationship between the original manufacturing share and the populist vote share. We also expect a positive relationship between the rate of change in manufacturing employment and the populist vote because a slower decline indicates a slow pace of restructuring and in turn, fewer local job opportunities for previous manufacturing workers. To depict the impact of economic crisis, CYCL, in our case the Great Recession, on the populist vote share, we use the rise of local unemployment rates until the peak of the recession and the reduction of unemployment after the recession as proxies. We would expect that regions where the recession had a stronger negative impact on employment (a faster rise in unemployment during the recession) and where labour markets recovered slowly from the recession (a slower rate of decline in unemployment rates) exhibit a higher popular vote share. The current economic situation, ECON, is represented through the current unemployment rate. Median household income is highly negatively correlated with the unemployment rate and could not be included. Ideally, we would also have liked to examine the impact of rising inequality on voting behaviour, but unfortunately, information on that was not available for the UK. Instead, we were able to examine the local impact of austerity programs in the UK (data for the US and Austria were not available). In general, we would expect higher unemployment rates and higher losses in benefits per person to result in more dissatisfaction with the status quo and therefore be positively related to populist vote shares. However, if it is true that voters trust established parties and candidates more on economic issues, a higher unemployment rate could also result in a lower share of the populist vote (Ivarsflaten, 2005). Finally, we also include age and gender controls as well as regional population size. We also included education and the percent of White British (for the UK) and White (for the US) in the basic model. However, the problem is the high correlation between the share of the highly educated and our economic and immigration target variables. Hence, we could not control for higher education and white ethnicity in our final models. To account for broad regional trends in voting patterns, we included a Wales and London dummy in the UK case and region dummies for eight of the nine census regions in the US case (we omitted a dummy for the mid-Atlantic states). All three elections built on earlier gains of populist agendas. In the UK and in Austria, the share of the populist right-wing agendas shifted dramatically upwards from previous gains of the FPÖ in Austria and UKIP in the UK. On the other hand, Trump, as a populist but Republican candidate, could rely on the Republican core vote and attack floating voters in specific states through populist messages. In a second step, we therefore explore how well our economic variables do to explain gains or losses in the populist vote through the following model:  Votediff=α0+β1IMM+β2STRUCT+β3CYCL+β4ECON+β5X+β6D+ε (2) where Votediff refers to the difference in regional vote shares between previous elections and the most recent elections. In the UK, we subtracted the share of the UKIP vote of the European 2014 Parliamentary Elections from the Brexit vote, in the US, we subtracted the share of the Romney vote in the 2012 US presidential elections from the 2016 Trump share, and in Austria, we subtracted the share of the FPÖ vote of the last general election from the share of the Hofer vote. We made a minor adjustment to model (1) to highlight recent economic conditions by replacing the historic manufacturing share and the rate of change with the current manufacturing share. This was motivated also by Trump’s success of targeting current manufacturing workers in the old industrial heartland of the US. Because data for the US are based on samples, we used population weights to minimise measurement error in the US case. The exact specification of our variables is summarised in Appendix A. Empirical results Before discussing our results, we offer a table with simple correlation coefficients between regional socio-demographic and economic characteristics identifying the core voter potential of populist right parties for the three countries. While populist parties tend to rest on relatively fluid and poorly defined social bases, Arzheimer (2011) argues that the electorate of the populist right in different countries nevertheless does consist of a clearly defined social core. They tend to be overwhelmingly male, poorly educated, ‘lower’ social classes (petty-bourgeoisie of artisans, shopkeepers, farmers and other self-employed are increasingly supplemented with non-traditional workers, members of the lower middle classes and the unemployed) and younger voters. Table 1 shows how the spatial distribution of those ‘core characteristics’ correlate with the populist vote shares in our three country cases. Because occupation is highly correlated with education and because this variable is not readily available on a consistent basis for the three countries, we omit occupation. We also add ‘ethnicity’ and population size of regions, as the strong nativist tone in the three elections should resonate more with the ‘native’, ‘pure’ population which is defined as white in the UK and US case. In Austria, we simply look at the share of foreign-born, as the share of immigrants that would fall under ‘non-white’ would be negligible. As the campaigns were also run on a traditional/cosmopolitan cleavage, we include regional size as we believe that large urban areas tend to be more cosmopolitan in outlook, while rural areas tend to be more inward looking and traditional. Table 1. Simple correlation coefficients between “core populist right voter characteristics” and populist vote shares.   Hofer %  Brexit %  Trump %  Young  −0.12  −0.30  −0.21  Old  +0.14  +0.41  +0.41  Male  +0.15  −0.20  +0.30  High education  −0.56  −0.89  −0.57  Medium education  n.a.  +0.78  +0.64  Low/no education  +0.30  +0.76  −0.07  Ethnicity  n.a.  +0.54  +0.67  Population size  −0.14  −0.18  −0.69  Number of regions  2122  348  3108    Hofer %  Brexit %  Trump %  Young  −0.12  −0.30  −0.21  Old  +0.14  +0.41  +0.41  Male  +0.15  −0.20  +0.30  High education  −0.56  −0.89  −0.57  Medium education  n.a.  +0.78  +0.64  Low/no education  +0.30  +0.76  −0.07  Ethnicity  n.a.  +0.54  +0.67  Population size  −0.14  −0.18  −0.69  Number of regions  2122  348  3108  Young: UK: % 18–24 year olds; US: % 20–24 year olds; AT: % 20–24 year olds; old: % >65 year olds; male: % of male population; high education: % with college degree or higher; medium education: UK: % with some education but no college degree; US: % with high school degree; low education: UK: % with no qualifications; US: % with less than high school degree; AT: % with apprenticeship or no formal qualification; ethnicity: UK: % white British; US: % white; population size: logarithm of total population. In the US, observations are weighted by population size. View Large The most striking result from Table 1 is the high correlation of the education variables with the Brexit vote and, to a lesser extent, the Trump and Hofer votes. In the UK, the share of college graduates correlates highly negatively with the Brexit vote. The negative coefficients are smaller for the Trump and Hofer votes. In the UK and US, the share of those with medium education levels is highly positively correlated with the Trump and Brexit vote shares, and the share of those with no or low qualifications is positively correlated with the populist vote in the UK and Austrian cases but insignificant in the US case. In all cases, the share of the young is negatively correlated with the populist vote share, while the share of the old is positively correlated. Correlations for the percentage of males are small and have different signs, which can be explained by the small spatial variation of this variable. The share of white British and Whites is strongly correlated with the Brexit and Trump vote share. In the US, the size of a county exhibits the highest correlation coefficient with the populist vote indicating that the ‘traditional/cosmopolitan’ distinction is particularly relevant there. Based on these simple correlations we could build a model around the relationship between education and the populist vote share. This would certainly yield a good model result, but what would it actually tell us? As the highly educated tend to occupy managerial and professional jobs that yield high salaries and whose jobs are less threatened by globalisation and immigration it seems plausible that they are less likely to vote for change or anti-globalisation or -immigration politics and policies. It could also be the case that the highly educated tend to be more liberal and hence, are less likely to vote for the authoritarian policies of the populist right. Or, at the level of regions, does it mean that their skills allow them to move to urban centres that tended to benefit from economic integration at the expense of rural areas and old industrial regions, and that these economic opportunities rather than individual attributes make them vote against populist agendas? In this case, the share of the highly educated correlates with a number of other regional characteristics that we would expect to influence voting outcomes. Indeed, if we look more carefully at the relationships between the share of the highly educated and our target variables, then the share of the highly educated is highly correlated with the share of foreign-born (and in the US with city size), but highly negatively correlated with unemployment rates or manufacturing employment shares in regions. Similarly, the share of white British and Whites is strongly negatively correlated with immigration and ethnic minority variables. In an ideal world, we would combine individual-level data with regional contextual data to isolate different processes. But as we do not have access to detailed individual-level data, we choose to model more explicitly the impact of regional economic context and immigration and explore how much those variables account for regional variation in the populist vote shares. Table 2 presents the results of a simple OLS regression, including robust standard errors and beta values. The dependent variable is the percentage of the populist vote. Parameter estimates for control variables age and gender as well as region dummies and the constants are omitted to allow a better focus on the target variables. The first observation is that the results are fairly robust across the three electoral decisions. The parameter estimates for the level of immigration (the share of foreign-born in Austria, the share of EU15 born in the UK, the share of Hispanics in the US) are negative and significant. Their relative effects are also high in all three countries. While the level of Eastern European accession country-born is insignificant in the UK, the increase in immigration (increase in foreign-born, increase in Eastern European accession state citizens, increase in Hispanics) is significant and positive in all three electoral outcomes, although the effect is substantially lower than for the negative impact of the share of immigrants. This result is consistent with Goodwin and Heath (2016), who identified the immigration accelerator effect for the UK. Table 2. Parameter estimates for Model (1), with robust standard errors in parentheses and beta values in brackets. Dependent variables    % Hofer  % Brexit  % Brexit  % Trump  Immigration  Level  −0.454**  −2.327**  −2.501**  −0.214**      (0.044)  (0.787)  (0.727)  (0.035)      [−0.287]  [−0.356]  [−0.382]  [−0.208]    EU accession countries, level    −0.328  −0.621          (0.337)  (0.320)          [−0.053]  [−0.100]      Change  0.002*  0.176**  0.187**  2.597**      (0.002)  (0.055)  (0.055)  (0.518)      [0.044]  [0.127]  [0.134]  [0.135]  Structural  Manufacturing %, base year  0.178**  0.269**  0.255**  0.343**      (0.023)  (0.055)  (0.054)  (0.037)      [0.190]  [0.239]  [0.227]  [0.261]    Manufacturing change  0.006  0.213**  0.185**  0.018**      (0.019)  (0.039)  (0.037)  (0.003)      [0.010]  [0.271]  [0.235]  [0.145]  Cyclical/recession  Impact  −0.011**  0.103**  0.110**  0.070**      (0.006)  (0.018)  (0.016)  (0.012)      [−0.066]  [0.237]  [0.253]  [0.202]    Recovery  0.015**  0.064  0.148**  0.131**      (0.007)  (0.046)  (0.036)  (0.049)      [0.060]  [0.052]  [0.120]  [0.069]  Current  Unemployment rate  0.545**  3.554**    −0.118      (0.120)  (0.501)    (0.343)      [0.171]  [0.317]    [−0.010]    Benefit loss per person      0.027**            (0.003)            [0.340]    Size  Log of population size  −0.009**  −2.581**  −2.802**  −4.093**      (0.003)  (0.664)  (0.612)  (0.460)      [−0.087]  [−0.147]  [−0.159]  [−0.383]  Regional dummies    No  Yes  Yes  Yes  Population weights    No  No  No  Yes  R2    0.19  0.72  0.74  0.69  Dependent variables    % Hofer  % Brexit  % Brexit  % Trump  Immigration  Level  −0.454**  −2.327**  −2.501**  −0.214**      (0.044)  (0.787)  (0.727)  (0.035)      [−0.287]  [−0.356]  [−0.382]  [−0.208]    EU accession countries, level    −0.328  −0.621          (0.337)  (0.320)          [−0.053]  [−0.100]      Change  0.002*  0.176**  0.187**  2.597**      (0.002)  (0.055)  (0.055)  (0.518)      [0.044]  [0.127]  [0.134]  [0.135]  Structural  Manufacturing %, base year  0.178**  0.269**  0.255**  0.343**      (0.023)  (0.055)  (0.054)  (0.037)      [0.190]  [0.239]  [0.227]  [0.261]    Manufacturing change  0.006  0.213**  0.185**  0.018**      (0.019)  (0.039)  (0.037)  (0.003)      [0.010]  [0.271]  [0.235]  [0.145]  Cyclical/recession  Impact  −0.011**  0.103**  0.110**  0.070**      (0.006)  (0.018)  (0.016)  (0.012)      [−0.066]  [0.237]  [0.253]  [0.202]    Recovery  0.015**  0.064  0.148**  0.131**      (0.007)  (0.046)  (0.036)  (0.049)      [0.060]  [0.052]  [0.120]  [0.069]  Current  Unemployment rate  0.545**  3.554**    −0.118      (0.120)  (0.501)    (0.343)      [0.171]  [0.317]    [−0.010]    Benefit loss per person      0.027**            (0.003)            [0.340]    Size  Log of population size  −0.009**  −2.581**  −2.802**  −4.093**      (0.003)  (0.664)  (0.612)  (0.460)      [−0.087]  [−0.147]  [−0.159]  [−0.383]  Regional dummies    No  Yes  Yes  Yes  Population weights    No  No  No  Yes  R2    0.19  0.72  0.74  0.69  Immigration level refers to percent of EU15 in UK, percent Hispanics in the US and percent foreign-born in Austria; change refers to rate of change in population from Eastern European EU accession countries in UK, difference in share of Hispanics in the US and rate of change of foreign-born in Austria; impact of recession refers to increase in unemployment rates prior to recession and recover refers to reduction in unemployment rate since the recession. ** (*) significant at 0.01 (0.05) level. View Large Old industrial areas, and especially those that have not reduced their manufacturing shares quickly enough, exhibit higher shares of the populist vote. Again, this result is consistent across all three countries (see also Shafique, 2016 for the UK). Moving on to the impact of the recession, we observe a positive relationship between the increase in unemployment rate going into the recession in the US and UK and, to a lesser extent, a positive relationship between slow recovery (unemployment rates decline at a slower pace) and the populist vote shares. In Austria, somewhat surprisingly, the increase in the local unemployment rates during the recession is associated negatively and a slow recovery is barely related to the Hofer share. This may have to do with the different evolution of Austrian unemployment rates. There, unemployment rates increased steadily until 2009, then dropped briefly until 2011, at which point they began to rise again. In the UK and US, the impact of the recession on the labour markets and its recovery from it were much more severe, and unemployment rates followed a clearer path of increase and decline prior and after the peak of the recession. Overall, the relative effect of the recession on the populist vote appears slightly less important than the historical importance of the manufacturing sector in a region. The final set of economic variables is the current unemployment rate and the impact of austerity measures (for the UK). In Austria and the UK, local, current unemployment rates have a big relative effect on populist vote shares, and this after controlling for the impact of the Great Recession. In the US, the current unemployment rate has no significant impact on the Trump vote share. This appears surprising, but may be explained by the fact that the core Republican voter is less concerned with unemployment and more with cultural and social policies. If this were the case, then unemployment should become important when explaining gains in the Trump over Romney vote. Because in the UK austerity measures were severe and varied substantially across localities, we also offer the results of a version that replaced the current unemployment rate with the average annual loss in benefits per person in each region.7 Apart from the share of EU15 born, the loss of benefits has the largest relative effect in the UK model. And finally, to account for a ‘cosmopolitan’ effect of cities that goes beyond the share of foreigners, we also included the population size as independent variable. This variable is significant and negative and particularly important to explain the Trump vote share in the US (this is the case for the weighted and unweighted models). While we do not dwell on the regional effects in the US elections, it is worth mentioning that the indicator variable for Wales was strongly negative and that for London positive. While the estimate for Wales does not come as a surprise, the positive value for London does. It suggests that given the low manufacturing share, lower unemployment rates and high share of immigrants, the Brexit voter share was, on average, 6.8% higher in London boroughs than we would expect given those conditions. The model fits for the US and the UK are reasonably high. For the Austrian case, the model fit is relatively poor. A population-weighted model would have pushed up the R2, but since we rely on census and not sample data, we did not consider that appropriate for the Austrian case. We now examine the enormous increase in the populist vote from previous elections (in the case of Austria and the UK) and to get a better understanding where and why Trump added to the Republican core vote. Table 3 provides the results of this analysis. Our dependent variables are now differences in the Hofer, Brexit and Trump vote shares from the FPÖ, UKIP and Romney vote shares in the last Austrian general elections, the 2014 EU parliamentary elections and the 2012 US presidential elections. Table 3. Parameter estimates for Model (2)   Austria  UK  US  Midwest  Immigration  Level  −0.487**  −1.425**  0.004  −0.148      (0.044)  (0.388)  (0.013)  (0.076)      [−0.326]  [−0.419]  [0.011]  [−0.185]    EU2    0.426            (0.263)            [0.132]        Change  0.003  −0.036  0.004  −0.266      (0.002)  (0.038)  (0.013)  (0.470)      [0.044]  [−0.050]  [−0.038]  [−0.027]  Structural  Manufacturing share, current  0.024  0.497**  0.097**  0.244**      (0.027)  (0.080)  (0.028)  (0.030)      [0.019]  [0.388]  [0.090]  [0.244]  Cyclical/recession  Impact  −0.010*  −0.008  0.003  −0.052**      (0.004)  (0.012)  (0.006)  (0.009)      [−0.060]  [−0.035]  [0.028]  [−0.229]    Recovery  0.021**  −0.080**  −0.026  −0.031      (0.006)  (0.030)  (0.019)  (0.022)      [0.088]  [−0.124]  [−0.041]  [−0.051]  Current  Unemployment rate  0.438**  2.578**  1.068**  1.869**      (0.111)  (0.335)  (0.190)  (0.147)      [0.146]  [0.443]  [0.285]  [0.397]  Size  Log of population size  −0.018**  −0.542  −1.132**  −1.394**      (0.003)  (0.436)  (0.133)  (0.199)      [−0.175]  [−0.059]  [−0.320]  [−0.398]  Regional dummies    No  Yes  Yes  No  Population weights    No  No  Yes  Yes  R2    0.30  0.61  0.60  0.74    Austria  UK  US  Midwest  Immigration  Level  −0.487**  −1.425**  0.004  −0.148      (0.044)  (0.388)  (0.013)  (0.076)      [−0.326]  [−0.419]  [0.011]  [−0.185]    EU2    0.426            (0.263)            [0.132]        Change  0.003  −0.036  0.004  −0.266      (0.002)  (0.038)  (0.013)  (0.470)      [0.044]  [−0.050]  [−0.038]  [−0.027]  Structural  Manufacturing share, current  0.024  0.497**  0.097**  0.244**      (0.027)  (0.080)  (0.028)  (0.030)      [0.019]  [0.388]  [0.090]  [0.244]  Cyclical/recession  Impact  −0.010*  −0.008  0.003  −0.052**      (0.004)  (0.012)  (0.006)  (0.009)      [−0.060]  [−0.035]  [0.028]  [−0.229]    Recovery  0.021**  −0.080**  −0.026  −0.031      (0.006)  (0.030)  (0.019)  (0.022)      [0.088]  [−0.124]  [−0.041]  [−0.051]  Current  Unemployment rate  0.438**  2.578**  1.068**  1.869**      (0.111)  (0.335)  (0.190)  (0.147)      [0.146]  [0.443]  [0.285]  [0.397]  Size  Log of population size  −0.018**  −0.542  −1.132**  −1.394**      (0.003)  (0.436)  (0.133)  (0.199)      [−0.175]  [−0.059]  [−0.320]  [−0.398]  Regional dummies    No  Yes  Yes  No  Population weights    No  No  Yes  Yes  R2    0.30  0.61  0.60  0.74  Dependent variables are the differences in vote shares from previous elections (UKIP vote for EU parliamentary elections in the UK, percent Romney vote in the US; FPÖ vote in Austria). Robust standard errors are in parenthesis and beta values in brackets. Immigration level refers to percent of EU15 in UK, percent Hispanics in the US and percent foreign-born in Austria; change refers to rate of change in population from Eastern European EU accession countries in UK, difference in share of Hispanics in the US and rate of change of foreign-born in Austria; impact of recession refers to increase in unemployment rates prior to recession and recover refers to reduction in unemployment rate since the recession. ** (*) significant at 0.01 (0.05) level. View Large Table 3 highlights a number of issues. The share of immigrants is negatively related to the populist vote share gains in Austria and the UK, but has no impact on Trump’s vote gain. The increase in immigration had no further impact on increasing the populist vote shares. Those for whom rising immigration was an issue voted populist prior to the recent electoral decisions, it appears. The share of current manufacturing employment is important in the UK (Shafique, 2016) and the US, in particular in the Midwest. In Austria, the impact of the manufacturing share is insignificant. The current unemployment rate is important to explain the populist vote share increase in all three elections, in particular the UK and the American Midwest. In other words, while the unemployment rate is not important to explain the basic Republican vote share, it is important to explain the gains Trump made over Romney. Population size is strongly negatively related to populist vote gains in Austria and the US but not in the UK. The impact of the recession on labour markets appears to have had little effect on increasing the populist vote share. Current unemployment rates are more important. Overall, the model performs best to explain the Trump gains in the Midwest, where his team targeted areas with sizeable manufacturing bases and where the message of protecting American interests and industries fell on fertile ground. Conclusion Drawing inspiration from macro-level studies on the role of modernisation processes and economic crisis on the rise of populist right parties, candidates and agendas, we examined the impact of subnational economic context on the regional share of the Hofer, Brexit and Trump votes. We believe that more attention to the subnational scale is important to obtain a tighter fit between the theoretical arguments linking the effects of socio-economic and cultural change to individual voting behaviour because modernisation processes in form of globalisation, deindustrialisation and migration affect regions and in turn, the environment in which individuals make political choices, differently. Aggregate studies at the national scale gloss over those differences. Our results demonstrate that economic variables perform reasonably well to explain subnational variation in the Brexit and Trump votes, but less so to explain regional variation in the Hofer vote. Regional shares in the populist right-wing vote tend to be higher in old industrial regions, are more pronounced in those that restructure more slowly, tend to be higher in areas with high current unemployment rates, those areas whose labour markets were affected more severely by the Great Recession, and they were lower in larger regions and those that were characterised by higher shares of foreign-born. Importantly though, regions where immigrant shares rose faster tended to exhibit higher shares of the populist right-wing vote. This may indicate that rapidly rising immigrant shares increase the demand for local services. Or, it may mean that austerity measures reduce the supply of local services and benefits that will be felt the most in regions with increased demand for these services. For the UK, we could show that faster increases in immigrants from EU accession countries and higher losses in benefits had a positive effect on the regional Brexit vote share. While these first-cut results are promising, the interaction between the supply of and demand for public services and welfare benefits and their effects on the populist right-wing vote share in different regions and countries needs further, detailed investigation. When switching the focus to increases in vote shares, we attempted to examine the impact on floating rather than core voters of populist candidates and parties. Here, current economic conditions seemed more important than the cyclical effects or increases in immigration. In particular, the current unemployment rate, the level of migration and, for the US, a high share in manufacturing employment was important to explain those vote increases. For the US, the share of manufacturing was a stronger explanatory factor in the Midwest than the country as a whole, which suggests that the supply of a populist agenda targeting disenfranchised manufacturing workers in the old industrial heartland was important for the outcome of the election. Not too surprisingly, there are a number of problems that affect almost all cross-sectional analyses. Without access to individual-level data, we are unable to establish causal links between our economic variables and individual voting behaviour. However, we also believe that the results of this exploratory analysis are encouraging and invite further work in the following directions: First, the individual country case studies need to be developed in greater detail drawing, for instance, on historical voting records and taking into account migration patterns to separate cause from effect: is it that individuals become more open to change as a result of their exposure to diversity and hence, are less likely to vote for populist right ideas, or are those that are more open to new ideas more likely to leave structurally declining areas and move to more diverse cities where they continue to vote against the populist right? Second and related, noting that parameter estimates are lower but in line with the US and UK case, what may explain the poor model fit for Austria? One explanation may be that the Austrian manufacturing and tourism sectors benefitted from joining the EU, that parts of the old nationalised metals and steel industries have been restructured successfully and that the small- and medium-sized enterprise sector has been thriving, at least prior to the recession. Because of the relatively high standard of living, even among the unskilled and less educated, and the lower levels of spatial and social inequality, economic explanations may be less successful in explaining the populist vote in Austria. One of the key groups of FPÖ voters are the ‘welfare chauvinists’ (Fallend, 2013), and it may be the fear of losing those welfare benefits to immigrants and refugees rather than the actual, local economic context that motivated Austrians to vote for Hofer. However, these hypotheses need to be explored and corroborated through further analysis. Third, as one of the key changes since the 1980s is increasing inequality and social polarisation, it would be important to examine whether inequality has an effect on the populist vote shares. Fourth, future work needs to look more carefully at interactions between different variables. For instance, what is the impact of the level and increase of immigration on the populist right-wing vote share in high and low unemployment areas, in fast and slow growing regions, in old industrial regions and new industrial spaces? Or, what is the impact of unemployment or low education levels on the populist vote shares in high growth and low growth regions? What is the impact of welfare losses in high and low immigration areas? Again, those questions are easier carried out for individual countries first. Acknowledgements We thank Maria Reyero for her assistance in compiling data for the US and Anna Stelzer for her assistance in compiling the data set for Austria. Data collection for the Austrian case was supported by the WU Small Scale Projects scheme, grant no. 11000416. We also would like to thank the editorial team for their support in turning around the script in record time. References Arzheimer, K. ( 2011) Electoral sociology: who votes for the extreme right and why – and when? In U. Backes and P. Moreau (eds) The Extreme Right in Europe: Current Trends and Perspectives , pp. 35– 50. Göttingen: Vandenhoeck and Ruprecht. Google Scholar CrossRef Search ADS   Autor, D., Dorn, D., Hanson, G. and Majlesi, G. ( 2016) A Note on the Effect on Rising Trade Exposure on the 2016 Presidential Elections . Available online at: https://gps.ucsd.edu/_files/faculty/hanson/hanson_research_TrumpVote-032017.pdf [Accessed 4 April 2017]. Becker, S. O., Fetzer, T. and Novy, D. ( 2016) Who Voted For Brexit? A Comprehensive District-Level Analysis . Coventry: Centre for Competitive Advantage in the Global Economy, Department of Economics, The University of Warwick. Working Paper Series, October 2016, No. 305. Betz, H. G. ( 1994) Radical Right-Wing Populism in Western Europe . New York, NY: St. Martin’s Press. Google Scholar CrossRef Search ADS   Blyth, M. ( 2016) Global Trumpism: Why Trump’s Victory Was Thirty Years in the Making and Why It Won’t Stop Here . Foreign Affairs, November 2016. Available online at: https://www.foreignaffairs.com/articles/2016-11-15/global-trumpism. Blyth, M. and Matthijs, M. ( 2017) Black swans, lame ducks, and the mystery of IPE’s missing macroeconomy, Review of International Political Economy , 24: 203– 231. Google Scholar CrossRef Search ADS   Eribon, D. ( 2009) Die Rückkehr nach Reims . Frankfurt: Suhrkamp. Esping-Anderson, G. ( 1990) Three Worlds of Welfare Capitalism . Princeton, NJ: Princeton University Press. Evans, J. A. J. ( 2005) The dynamics of social change in radical right-wing populist party support, Comparative European Politics , 3: 76– 101. Google Scholar CrossRef Search ADS   Evans, G. and Mellon, J. ( 2016) Working class votes and conservative losses: solving the UKIP puzzle, Parliamentary Affairs , 69: 464– 479. Google Scholar CrossRef Search ADS   Fallend, F. ( 2013) Populism in government: the case of Austria (2000–2007). In C. Mudde and R. Kaltwasser (eds) Populism in Europe and the Americas , pp. 113– 135. Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Ford, R. and Goodwin, M. ( 2017) A nation divided, Journal of Democracy , 28: 17– 30. Google Scholar CrossRef Search ADS   Golder, M. ( 2003) Explaining variation in the electoral success of extreme right parties in Western Europe, Comparative Political Studies , 36: 432– 466. Google Scholar CrossRef Search ADS   Golder, M. ( 2016) Far right parties in Europe, Annual Review of Political Science , 19: 477– 497. Google Scholar CrossRef Search ADS   Goodwin, M. J. and Heath, O. ( 2016) The 2016 referendum, Brexit and the left behind: an aggregate-level analysis of the result, The Political Quarterly , 87: 323– 332. Google Scholar CrossRef Search ADS   Harris, R. and Charlton, M. ( 2016) Voting out of the European Union: exploring the geography of leave, Environment and Planning A , 48: 2116– 2128. Google Scholar CrossRef Search ADS   Harvey, D. ( 2005) A Brief History of Neoliberalism . Oxford: Oxford University Press. Hobolt, S. B. ( 2016) The Brexit vote: a divided nation, a divided continent, Journal of European Public Policy , 23: 1259– 1277. Google Scholar CrossRef Search ADS   Inglehart, R. F. ( 1990) Cultural Shift in Advanced Industrial Society . Princeton, NJ: Princeton University Press. Inglehart, R. F. ( 1997) Modernization and Postmodernization: Cultural, Economic and Political Change in 43 Societies . Princeton, NJ: Princeton University Press. Inglehart, R. F. and Norris, P. ( 2016) Trump, Brexit, and the Rise of Populism: Economic Have-nots and Cultural Backlash . Faculty Research Working Paper Series. RWP16-026. August. Cambridge, MA: Harvard Kennedy School. Ivarsflaten, E. ( 2005) The vulnerable populist right parties: no economic realignment fuelling their electoral success, European Journal of Political Research , 44: 465– 492. Google Scholar CrossRef Search ADS   Johnson, P. ( 2015) Older Industrial Britain Being Left Behind. Available online at: https://newstartmag.co.uk/articles/older-industrial-britain-being-left-behind-in-economic-upturn/ [Accessed 3 September 2017]. Kriesi, H. P., Grande, E., Lachat, R., Dolezal, A., Bornschi, S. and Frey, T. ( 2006) Globalization and the transformation of the national political space: six European countries compares, European Journal of Political Research , 45: 921– 956. Google Scholar CrossRef Search ADS   Kriesi, H. P., Koopmans, R., Duyvendak, J. W. and Giugni, M. G. ( 1995) New Social Movements in Western Europe: A Comparative Analysis . Minneapolis, MN: University of Minnesota Press. Lipset, S. M. ( 1960) Political Man: The Social Basis of Politics . New York, NY: Doubleday. Los, B., McCann, P., Springford, J. and Thissen, M. ( 2017) The mismatch between local voting and the local economic consequences of Brexit, Regional Studies , 51: 786– 799. Google Scholar CrossRef Search ADS   Lubbers, M., Gijsberts, M. and Sheepers, P. ( 2002) Extreme right-wing voting in Western Europe, European Journal of Political Research , 41: 345– 378. Google Scholar CrossRef Search ADS   McCarty, N., Poole, K. T. and Rosenthal, H. ( 2016) Polarized America . Cambridge, MA: The MIT Press. Minkenberg, M. ( 2000) The renewal of the radical right: between modernity and anti-modernity, Government Opposition , 35: 170– 188. Google Scholar CrossRef Search ADS   Mouffe, C. ( 2016) The Populist Moment . Available online at: https://www.opendemocracy.net/democraciaabierta/chantal-mouffe/populist-moment [Accessed 11 January 2017]. Mudde, C. ( 2007) Populist Radical Right Parties in Europe . New York, NY: Cambridge University Press. Google Scholar CrossRef Search ADS   Mudde, C. and Kaltwasser, R. ( 2013) Populism and (liberal) democracy: a framework for analysis. In C. Mudde and R. Kaltwasser (eds) Populism in Europe and the Americas , pp. 1– 26. Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Müller, W. C. ( 2002) Evil or the ‘engine of democracy’? Populism and party competition in Austria. In Y. Meny and Y. Surel (eds) Democracies and the Populist Challenge , pp. 155– 175. Basingstoke: Palgrave. Google Scholar CrossRef Search ADS   Müller, J.-W. ( 2016) Was ist Populismus?  Berlin: Suhrkamp. Norris, P. and Inglehart, R. J. ( 2009) Cosmopolitan Communications . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Oliver, J. E. and Rahn, W. M. ( 2016) Rise of the Trumpenvolk: populism in the 2016 Election, The Annals of the American Academy of Political and Social Science , 667: 189– 206. Google Scholar CrossRef Search ADS   Piketty, T. ( 2014) Capital  in the 21st Century. Cambridge, MA: Belknap Press. Pool, K. and Rosenthal, H. ( 2009) Ideology and Congress . New Brunswick and London: Transaction Publishers. Shafique, A. ( 2016) BREXIT Was Driven by Places ‘Left Behind’. Available online at: https://www.thersa.org/discover/publications-and-articles/rsa-blogs/2016/08/brexit-was-driven-by-places-left-behind [Accessed 2 September 2017]. Sheppard, E. ( 2015) The Limits to Globalization . Oxford: Oxford University Press. Singh, R. ( 2017) ‘I, the people’: a deflationary interpretation of populism, Trump and the United States constitution, Economy and Society , 46: 20– 42. Google Scholar CrossRef Search ADS   Taggart, P. ( 2000) Populism . Buckingham: Open University Press. Tyson, A. and Maniam, S. ( 2016) Behind Trump’s Victory: Divisions by Race, Gender, Education . Pew Research Center. Available online at: http://www.pewresearch.org/fact-tank/2016/11/09/behind-trumps-victory-divisions-by-race-gender-education/ [Accessed 10 April 2017]. Endnotes 1 Although some may argue that there are also populist left-wing parties, in the North American and European context, they do not tend to pursue a strong nativist agenda and are not authoritarian. Also, in our case studies, we focus on three populist right-wing candidates and agendas. 2 Inglehart and Norris (2016) show that the average vote share of Western European populist parties in European parliamentary and national elections more than doubled since the 1960s, from 5.1 to 13.2%, at the expense of the centre parties. During the first round of the Austrian presidential elections, the two centre party candidates achieved a combined vote share of only 22.4% and were both eliminated from the run-off elections. 3 Notice though that Bernie Sanders in the US and Jeremy Corbyn in the UK had some electoral success when attempting to spell out those links, such that a failure by traditional parties to engage in this discourse is in part to blame for the simple “Eliminate Unemployment: Stop Immigration” (Golder, 2003: 438) slogan. 4 Officially, Alexander Van der Bellen ran as independent candidate. However, he was a long-standing leading figure of the Austrian Green Party in the past. 5 In the national legislative elections that took place on 15 October 2017, the FPÖ ended up with 26.0% of the vote and is currently in negotiation to form the next government with the conservative ÖVP whose new leader Sebastian Kurz delivered victory (31.5% of the vote) by adopting many of the anti-immigrant elements of the FPÖ’s party programme. The left-centre SPÖ, with 26.9% the second strongest party, will form the parliamentary opposition. 6 Although we would have preferred data on manufacturing employment for 1980, they were not available at municipality level for Austria. 7 Because of the high correlation between the current unemployment rate and benefit losses, we could not include both variables in the same model. Appendix A     Austria  UK  USA  Immigration  Level  Share of foreign-born, 2015  Share of EU15 born, 2011  Share of Hispanics, 2015    EU2    Share of Eastern European EU accession state citizens, 2011      Change  Rate of change in foreign-born, 2001–2015  Rate of change in EU accession citizens, 2001–2011  Difference in Hispanic share, 2005–2015  Structural  Manufacturing %, base year  Share of manufacturing employment, 1991  Share of manufacturing employment, 1981  Share of manufacturing employment, average 1979–1981    Manufacturing change  Rate of change in manufacturing employment, 1991–2015  Rate of change in manufacturing employment, 1981–2011  Rate of change in manufacturing employment, 1980–2015    Manufacturing share, current  Share of manufacturing employment, 2015  Share of manufacturing employment, 2011  Share of manufacturing employment, 2015  Cyclical/ recession  Impact  Rate of change in unemployment, 2001–2009  Rate of change in unemployment, 2001–2011  Rate of change in unemployment, 2005–2010    Recovery  Rate of change in unemployment, 2009–2015  Rate of change in job seekers allowance claimants, 2001–2016  Rate of change in unemployment, 2010–2015  Current  Unemployment rate  Unemployment rate, 2015  Job claimant rate, 2016  Unemployment rate, 2015    Benefit loss per person    Average annual loss in total benefits since benefit reform in 2014    Size  Log of population size  Log of total population, 2015  Log of total population, 2011  Log of total population, 2015      Austria  UK  USA  Immigration  Level  Share of foreign-born, 2015  Share of EU15 born, 2011  Share of Hispanics, 2015    EU2    Share of Eastern European EU accession state citizens, 2011      Change  Rate of change in foreign-born, 2001–2015  Rate of change in EU accession citizens, 2001–2011  Difference in Hispanic share, 2005–2015  Structural  Manufacturing %, base year  Share of manufacturing employment, 1991  Share of manufacturing employment, 1981  Share of manufacturing employment, average 1979–1981    Manufacturing change  Rate of change in manufacturing employment, 1991–2015  Rate of change in manufacturing employment, 1981–2011  Rate of change in manufacturing employment, 1980–2015    Manufacturing share, current  Share of manufacturing employment, 2015  Share of manufacturing employment, 2011  Share of manufacturing employment, 2015  Cyclical/ recession  Impact  Rate of change in unemployment, 2001–2009  Rate of change in unemployment, 2001–2011  Rate of change in unemployment, 2005–2010    Recovery  Rate of change in unemployment, 2009–2015  Rate of change in job seekers allowance claimants, 2001–2016  Rate of change in unemployment, 2010–2015  Current  Unemployment rate  Unemployment rate, 2015  Job claimant rate, 2016  Unemployment rate, 2015    Benefit loss per person    Average annual loss in total benefits since benefit reform in 2014    Size  Log of population size  Log of total population, 2015  Log of total population, 2011  Log of total population, 2015  View Large © The Author(s) 2018. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved. For permissions, please email: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Cambridge Journal of Regions, Economy and Society Oxford University Press

The victims of neoliberal globalisation and the rise of the populist vote: a comparative analysis of three recent electoral decisions

Loading next page...
 
/lp/ou_press/the-victims-of-neoliberal-globalisation-and-the-rise-of-the-populist-EwpObNhfgc
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved. For permissions, please email: journals.permissions@oup.com
ISSN
1752-1378
eISSN
1752-1386
D.O.I.
10.1093/cjres/rsx025
Publisher site
See Article on Publisher Site

Abstract

Abstract Recent presidential elections in the US and Austria as well as the referendum on Brexit in the UK delivered victories or near-victories for populist right-wing candidates or agendas. In all three cases, globalisation and European integration were blamed for higher immigration and pressure on public services, deindustrialisation and job losses, and the attack on traditional values by cosmopolitan elites supported by traditional centre parties that have been unable or unwilling to control those processes. While election analysts seek to explain voting behaviour with socio-demographic characteristics of individuals, individual voting preferences also depend on the geographical context in which decisions are made. This article thus examines how long-term, regional structural economic changes, the varying impact of the Great Recession on the rise of and recovery from regional unemployment and current regional economic conditions, such as unemployment and welfare benefit losses, affect regional vote shares. In addition to those economic conditions, we examine the impact of immigration and urban size on populist vote shares. We show that regions with low, but rising immigrant shares, old industrial regions, smaller regions, those whose labour markets were exposed more and recovered less from the Great Recession, and those with high unemployment rates and benefit losses exhibited higher populist vote shares. These results are largely consistent across the three case study countries. Introduction While populist parties did not suddenly burst on to the political scene after the Great Recession in the 2010s, a number of recent electoral decisions in the US and Europe returned wins or near-wins for right-wing populist candidates and their agendas. Many of the recently successful right-wing parties with a strong emphasis on ‘nativism’ and populism were either founded or moved from a liberal economic focus to a nativist focus in their policies in the 1980s or early 1990s (Mudde, 2007). While economic policies do not make up the core of populist party programmes, the rise of the populist vote nevertheless coincides with the rise of neoliberal globalisation and its impact on employment and wage structures in the countries of the Global North, coupled with an abandonment of the traditional working class and working class areas by traditionally left-wing parties (Eribon, 2009). This representative void may explain the shift in the communist vote in France or the Socialist vote in Austria to the populist right for instance. In addition to those gradual shifts towards the populist vote, we also witnessed a recent surge in the level of that vote relative to previous elections, which may be related, in part, to recent economic hardship produced by the Great Recession. A large number of competing and complementary conceptual frameworks have been employed to explain the rise of populist parties offering macro-, meso- and micro-level explanations of the demand for and supply of populist agendas, as well as the constitutional arrangements governing the electoral process (Golder, 2016; Mudde, 2007). These studies tend to focus on cross-country comparisons using individual survey data. However, while cross-country studies can highlight the role of broad economic, institutional and historical differences that may influence voting behaviour, and studies focusing on micro survey data can tell us something about the stated preferences of individuals, studies at the national scale do not account for pronounced geographical differences in economic, social and cultural contexts that generate demand for the populist vote (Golder, 2016). On the other hand, analysis of individual survey data tends to assume preferences and psychological profiles of respondents to be exogenous to the context in which those preferences are stated. Recent electoral outcomes involving populist parties or agendas were characterised by substantial geographical differences in the shares of the populist vote. By studying the geographical pattern of three electoral decisions, the presidential elections in the US and Austria as well as the Brexit vote in the UK, we want to explore whether geographic differences in selected demand factors for populist agendas offer reasonable explanations of electoral outcomes in those three countries. Focusing on the subnational rather than national scale of analysis should produce a closer matching of the actual opportunities and constraints that individuals face as a result of neoliberal globalisation and increasing economic and political integration, as well as the geographically varying impact of the Great Recession. While we do not claim to provide a comprehensive explanation of the rise of populist parties and candidates in Europe or the US, we offer a first-cut analysis of the role of regional economic differences to explain the populist vote shares, and through that flagging the potential importance of subnational electoral studies to political scientists. While they recognise heterogeneity among the preferences of individual voters, they ignore that those preferences are driven by growing up in families whose political preferences are shaped, in part, by socio-economic opportunities and constraints, coupled with the formation of cultural practices offered by the places in which they live. In our analysis, we thus focus on demand-side explanations of the populist vote, specifically factors that reflect the economic conditions of regions within the three countries we study. While most studies focus on unemployment rates, income or occupation to operationalise economic conditions, we distinguish between structural and cyclical changes as well as current economic conditions. Long-term structural decline as a result of increasing global competition and integration should result in a gradual increase in demand for populist agendas in those regions. We consider old industrial and mining regions to fall into that category. But voting decisions will also be influenced by cyclical factors, in particular the differential impact of the Great Recession on employment opportunities. And finally, voting should be affected by the current economic situation that individuals face. In addition, because all three electoral campaigns were fought on the back of ‘immigration crises’ and the lack of national control of it, we also examine the importance of migration to explain the share of the populist vote. More specifically, we ask the following questions: First, despite different national economic contexts, does the negative impact of neoliberal globalisation and tighter economic integration on old industrial areas result in higher populist vote shares? Second, to what extent does the impact of the Great Recession on regional employment affect the populist vote share? Third, how do the current regional economic conditions in terms of unemployment and welfare losses affect the populist vote share? Fourth, do regional differences in immigration rates influence the populist vote share? As our study is based on the cross-sectional data, we can explore how well our economic variables explain regional voting patterns, but are limited in establishing clear causal links between economic macro patterns and individual decision making. The article is structured as follows. The second section discusses the theoretical arguments for the rise of populism and their link to economic geography, while the third offers a brief overview of the gradual successes of populist parties and candidates in the three countries. Section four explains how the theoretical arguments are implemented empirically through a discussion of methods and data, the fifth section presents the main findings and section six concludes the paper. Populism and economic geography ‘Neoliberal globalisation’ and ‘populism’ are overused and conceptually stretched terms, but we use them to organise our empirical analysis and potential explanations of recent electoral decisions. Under ‘neoliberal globalisation’ we understand a specific form of globalisation in which the benefits of market integration become unquestioned truth (Harvey, 2005; Mouffe, 2016; Müller, 2016; Sheppard, 2015). In reality, the benefits and costs of globalisation are distributed unequally between social groups and different territories and so open up a cleavage between those in favour of more and those in favour of less integration (Kriesi et al., 2006). Political elites do not question seriously the perceived positive effects, thus those bearing the cost feel increasingly misunderstood, unheard and unrepresented by those political elites. While neoliberal globalisation does not cause the rise of populism, the undifferentiated belief in the benefits of globalisation by traditional parties on the right and left of the political spectrum and lack of political representation of those harmed by the process generates a representational void that is increasingly filled by populist parties, even if economic policy is not generally part of their core programmes. To understand why globalisation may result in the rise of populism we need a working definition of ‘populism’. Populist right-wing parties, candidates and their agendas1 have three characteristics. They are (i) anti-establishment, (ii) nativist, and often (iii) authoritarian. The populist right-wing vote is directed against perceived corrupt elites who are supposedly not representing the will of ‘the people’, understood by populists as a homogenous group of nationals with the exclusive rights to national resources (Müller, 2016). Because of the focus on ‘the pure people’ and the exclusive rights to resources attributed to them, populists are also anti-pluralist. They reject the need to compromise in order to reconcile differing group interests (Mudde and Kaltwasser, 2013). In those respects, a number of recent elections in the UK, US, Austria, the Netherlands, France, Hungary or Poland can therefore be interpreted as mandates for populist right-wing candidates promising to keep others out, taking control over their own resources and getting one over on the elites that supposedly do not know what the real people want and need. The question is what factors and processes explain the rise of right-wing populism?2 Here we briefly review some of the most important theories and discuss how they may help us understand geographic variation in the populist vote. Broadly, explanations can be separated into demand for populist parties, the supply of populist political programmes, messages and parties and the institutional context in which those parties operate (Mudde, 2007). The focus in the empirical literature on populism is on demand-based explanations that can broadly be grouped into those emphasising the role of economic and those emphasising the role of cultural changes as main cause for the rise of demand for populist agendas (Inglehart and Norris, 2016). One influential argument is based on the idea that recent developments and processes such as globalisation, deindustrialisation, rising inequality or the transition to post-Fordism generated a new group of Modernisierungsverlierer (losers of modernisation). While echoing Lipset’s (1960) analysis of the 1930s, where rapid industrialisation generated fears of downward social and economic mobility among the petit-bourgeoisie, small entrepreneurs, shopkeepers, merchants, self-employed artisans and independent farmers squeezed between big business and organised labour, that made them susceptible to fascist and extremist ideas, recent changes generated a new underclass that forms a new electoral base of populist right-wing parties (Betz, 1994; Esping-Anderson, 1990). While continuing to appeal to parts of the petite-bourgeoisie, populist right-wing parties now find new demand among the low-skilled, poorly educated, blue-collar working class whose jobs have been eradicated by globalisation, automatisation, deindustrialisation, have become more contested through competition from low-skilled immigration or devalued through stagnating or declining wages exacerbated by blue-collar union decline, austerity policies and an increasing inability of nation states to regulate big business or immigration (Betz, 1994; Inglehart and Norris, 2016; Piketty, 2014). These economic changes are translated into a new cleavage structure (Kriesi et al., 1995). In the past, European party systems have been dominated by economic (class) cleavage. Globalisation and deindustrialisation have led to a decline in class voting and partisan identification, increased political alienation among certain segments of the population and reduced trust in the political elite (Golder, 2016, 488). Changes in the occupational structure and resulting social group affiliation, as well as increasing secularisation on the centre-right, eroded the traditional voter pool, which meant that traditional centre and left parties searched for voters elsewhere and abandoned, to a large extent, their traditional base (Eribon, 2009). This process increased electoral volatility and created a pool of floating voters rather than traditional partisan loyalties generating a demand for representation that populist parties were able to capture (Betz, 1994; Minkenberg, 2000). Harnessing support from an economically diverse group of small business owners, white-collar workers, skilled and unskilled blue-collar workers, as well as the unemployed, necessitated identification through cultural, not economic values (Evans, 2005). The cultural cleavage that is forged rests on group identities based on nationality and ethnicity. Ethnic minorities and immigrant groups are blamed for deteriorating conditions, loss of jobs and increasing pressure on welfare services. While the literature on Modernisie rungsverlierer focuses on the long-term impact of economic changes on voting behaviour, a second set of work focuses on the role of crisis (Kriesi et al., 1995; Taggart, 2000). During economic crises the pressure on and competition for jobs and welfare services increases and hence, discussions about how those resources should be allocated among different groups become even more important. Immigrants and ethnic minorities are easier to blame than complex relationships between technological change, globalisation, neoliberal austerity policies and resource constraints.3 If we were to follow the logic of this argument, we would expect increasing vote shares for populist candidates during economic crisis and in regions affected most negatively by a crisis. However, the empirical evidence is mixed. While Lubbers et al. (2002) find a positive relationship between unemployment and voting for populist parties at the individual level, they fail to identify the link at the macro level. Theoretically, it is possible to argue that during times of crisis, voters fall back on traditional parties, as they are trusted more on economic issues than populist parties who mobilise their voters along cultural commonalities (Ivarsflaten, 2005). Golder (2003) established a more direct link between crisis, unemployment and immigration, and finds that unemployment is positively related to populist vote shares in countries with high immigration but less so if immigration is low. The idea of cultural cleavage as reaction to socio-economic changes induced by economic restructuring processes since the 1970s is picked up and re-formulated more broadly as ‘cultural backlash thesis’ by Inglehart and Norris (2016). They interpret the rise of the populist vote as cultural counter-revolution against the rise of progressive values, including increased tolerance towards sexuality, same-sex marriage, LGBT rights, ethical norms, open-mindedness towards migrants, refugees, foreigners, multicultural lifestyles, foods and travel, cosmopolitan support for international cooperation, humanitarian assistance, and multilateral agencies like the UN or EU, environmental concerns, race and gender equality and human rights (Inglehart, 1990, 1997; Norris and Inglehart, 2009). The transition from materialist to post-materialist values is accomplished through generational replacement and is linked to the rise of Green parties as well as progressive social movements and transnational activist organisations. As the share of post-materialists increases in the population, they brought new issues into politics, resulting in a declining emphasis on social class and redistribution, and an increased polarisation around cultural issues and social identities. Those favourable towards progressive social change and humanistic values tend to be economically secure and better educated. The rise of those values in the 1960s and 1970s was opposed early on by older traditionalists who felt threatened by the erosion of values. In the context of Western European societies, it is relevant that younger generations tend to grow up with greater economic security, and people growing up in secure circumstances have been shown to be more open-minded, socially tolerant and trusting, more secular and accepting of diversity. In addition, increasing participation of women in the workplace and higher education altered traditional gender roles. Traditional values were therefore held increasingly by the old, men and uneducated. In this view, xenophobia and hostility and intolerance towards migrants, ethnic and racial minorities are only part of a wider cultural backlash against social and cultural change. Empirically ‘culture’ is measured as educational attainment, trust in governance, anti-immigrant sentiments, authoritarian values or right-wing ideology (Inglehart and Norris, 2016). But scores on those variables are not independent of the economic and social positions of the respondents or the positions of families and/or regions and neighbourhoods they grew up in. A wider problem is that empirical results produce contradictory evidence for both the economic and cultural theses. In part, this can be explained by the difficulty in isolating cultural from economic effects, but it may also result from cross-sectional analysis of national election results. Considering substantial regional variation in electoral outcomes (Becker et al., 2016; Golder, 2016), focusing on whole countries appears problematic. For instance, while we may obtain a positive correlation between the share of immigrants and populist party vote at the national scale this does not have to be the case at the subnational scale, if immigrants are concentrated in large metropolitan areas that, overall, benefit from globalisation and economic integration, while rural areas (with very few immigrants) suffer. In this case, the relationship between immigration and populism may turn negative at the subnational scale. Analysing three electoral outcomes at the subnational scale should generate a tighter fit between the theoretical arguments linking economic, social and cultural contexts and voting outcomes. There is now some literature on regional differences in the Brexit vote (Becker et al., 2016; Goodwin and Heath, 2016; Harris and Charlton, 2016; Hobolt, 2016; Los et al., 2017), but this work tends to focus on socio-demographic variables, psychological variables, immigration, survey data or current economic conditions and neglects the role of long-term structural decline or the impact of and recovery from the Great Recession as potential explanatory variables for regional Brexit vote shares. For the US, Autor et al. (2016) examine the potential role of import competition from China on the increase in the Trump vote share, but they do not further examine long-term structural decline and cyclical effects as potential causes of higher Trump vote shares. We do not know of any studies that attempt to explain regional variation in the outcome of the Austrian presidential elections yet. Rather than restricting our analysis to one country, the comparison between three elections allows us to examine whether subnational economic changes affect the electoral outcomes in similar ways, despite differences in national political landscapes, differential exposure to and impacts of modernisation processes on the economies, sectors and occupational structures of those countries. While the Brexit vote and the Austrian presidential elections were decided by vote shares of the national electorate, the US presidential election depends on winning the electoral college of individual states. In this sense, supply-side factors, that is party strategies to shift non-Republican votes, need to target particular groups of voters in states that can be turned by them. We explore this by conducting separate analysis for individual regions within the US. Examining changes in populist vote shares allows us to examine whether regional differences in the long-, medium- and short-term economic conditions have a stronger impact on the populist electoral base or their floating voter shares (Arzheimer, 2011). We engage with and add to this literature in the following way: First, we add three empirical studies at the subnational scale to the sparse empirical literature on regional differences in populist vote shares. This is important because the large geographical variation in populist vote shares will cause problems for national-scale analysis if the postulated roots of increasing populism such as globalisation or deindustrialisation affect regions and people living on those regions differently. Second, attempting to explore regional voting patterns in three countries through identical model specifications allows us to examine whether regional context has similar effects on voting patterns, despite national historical and institutional differences. Third, in the empirical analysis below we favour economic over cultural explanations of rising populism. Although we readily admit that the rise of populism cannot be reduced to economics and will always have a social and cultural dimension, we follow Eribon (2009) and believe that the structural position of individuals in a socio-economic system will determine, in part, their social group membership and associated group-specific cultural practices, so that the influence of culture and social group membership cannot be independent of and isolated from economic context in which individuals grow up. Also, because our empirical work is based on cross-sectional data and ‘cultural’ and ‘economic’ variables correlate highly at the regional scale, we have little choice but to focus on one of the two. Fourth, we separate empirically the impact of modernisation and economic crisis on rising populist voter potential by distinguishing between long-term economic changes, the effects of the Great Recession and current economic conditions in a region. Trends of populist right voting in the three countries To set the scene, we provide some descriptive historical and geographical information on the populist vote shares in those countries. In the UK, Brexit was strongly pushed by the UK Independence Party (UKIP), in the US, the Republican Party and Tea Party moved increasingly towards the right and in Austria, the Freedom Party (FPÖ) enjoyed considerable vote gains since the election of Jörg Haider as their party chairman in 1986. Because of the first past the post system of the UK electoral system, UKIP only managed one seat in the House of Commons in the 2015 general election. However, this masked the rise of UKIP as an electoral force in the UK. Founded in 1993, UKIP increased its percent of overall votes in general elections from 0.3 in 1997 to 12.6 in 2015. At the same time, it rose to become the strongest British faction in the European parliament, where they obtained 27.5% of the vote and 24 of the 73 seats to be won in the 2014 EU parliamentary election. They were able to gain votes from an increasingly disenfranchised (and socially conservative and inward looking) working class neglected by New Labour and social conservatives on the right. The accumulated pile of populist votes was set on fire by increasing concerns over immigration from the EU accession countries ignited by Tony Blair’s decision not to follow other EU countries and pose restrictions on immigration from Eastern European EU countries, and kindled by David Cameron’s conservative party by promising to reduce net migration to ‘tens of thousands’ prior to the 2010 elections (Evans and Mellon, 2016; Ford and Goodwin, 2017). While a large share of UKIP voters comes from a conservative base, Brexit campaigning by UKIP tended to concentrate on the old industrial, labour voting areas and the workers left behind by globalisation and integration. Coupled with a lacklustre commitment of Labour Party leader Jeremy Corbyn to Europe, old industrial areas then voted overwhelmingly for Brexit (Goodwin and Heath, 2016; Johnson, 2015; Shafique, 2016). Although there are some question marks whether Donald Trump can really be seen as populist (Oliver and Rahn, 2016) or not (Singh, 2017), roll call vote studies show increasing polarisation and shift towards the right of the US Congress and Senate since the late 1970s (McCarty et al., 2016; Pool and Rosenthal, 2009). In that sense, Trump built on a right-moving Republican base and harnessed growing sentiments against the effects of neoliberal globalisation, high debt, low employment and wages and rising inequality by blaming the uncaring and unknowing elites, as well as easy scapegoats (Mexican immigrants and Chinese exports) packaged into simple slogans that pitch the true Americans against the rest of the world (Blyth, 2016; Blyth and Matthijs, 2017). During the election, the usual Republican voters turned up, but Trump did especially poorly with women and those with college education, did better than previous Republican candidate Mitt Romney with minorities, and did exceptionally well with white voters without a college degree, in particular in the old industrial heartland of the Midwest and, to a much lesser extent, among the manufacturing workers in the South (relative to Romney’s vote shares) (Tyson and Maniam, 2016). In the end, Trump managed to deliver 304 electoral votes to Clinton’s 227 despite losing the popular vote by approximately 3 million votes (46.1% compared to Clinton’s 48.2%). It was the capture of a large number of states in the old rustbelt of the US that included the turning of the blue states of Michigan, Pennsylvania and Wisconsin and the swing states of Iowa and Ohio that delivered the presidency. In Austria, although the President is only the head of state with de facto limited powers, the presidential election was seen as test case for the upcoming general elections in October 2017. As in the UK and the US, the first round of the election highlighted voter dissatisfaction with the status quo represented by the two centrist parties, the SPÖ (Social Democratic Party of Austria) and ÖVP (the conservative Austrian People’s Party), whose candidates obtained 11.3% (SPÖ) and 11.1% (ÖVP) of the vote, respectively. Norbert Hofer, the candidate of the populist right, the Austrian Freedom Party (FPÖ), won the first round with 35.1%, followed by the independent candidate,4 Alexander Van der Bellen, with 21.3%. Similar to the French case, voters from all parties rallied to avoid the populist right-wing candidate from winning in the run-off elections. Even so, he still managed to obtain 49.65% of the vote. Because of voting irregularities, the result of the first run-off election was annulled and in the re-run, Hofer obtained 46.21%. Hofer was able to build on the success of the FPÖ in national and regional politics since the party was transformed by Jörg Haider in 1986 into a right-wing populist party opposing the elites represented by the two centrist parties that dominated post-WWII Austrian politics, ditching former neoliberal economic and German national roots from the official party programme and replacing it with anti-immigrant, anti-EU, anti-globalisation, anti-consumerist positions to defend ‘true’, national Austrian values based on Christian heritage, law and order and welfare benefits (welfare chauvinism) from Islamic fundamentalism, immigrants and aggressive capitalism (Fallend, 2013; Müller, 2002). As a result, the FPÖ was able to increase its vote share in national elections from 5.0% in 1983 to 26.9% in 1999 and was able to win the provincial election in the Southern state of Carinthia, where it replaced the SPÖ as leading party.5 It is important to notice that this transformation and exploitation of the cultural cleavage between traditionalist and cosmopolitans was made possible by rapidly deteriorating economic conditions in the early 1980s, exemplified by the jump in unemployment rates from 1 to 4% in 1982, reflecting job losses in the nationalised steel and metal industries. This step-change in unemployment demonstrated the inability of centrist parties to deliver full employment and maintain a high-level welfare state in the face of neoliberal globalisation. While the exploitation of cultural cleavages emerges as key element in all three electoral decisions, those cleavages opened up as a result of deteriorating economic conditions for traditional blue-collar workers that now supplement the original voter potential of the right that consisted of nationalists and the petite-bourgeoisie. But in addition to those gradual shifts, the recent electoral outcomes represented a further dramatic shift towards the populist right (especially in the UK and Austria) that dovetail with further deteriorating economic conditions and austerity policies brought on by the Great Recession. The literature on the rise of populism is rather silent on those possible impacts of short-term economic changes on the populist vote share. In addition, while explanations based on traditional–cosmopolitan cleavages should be reflected in simple rural–urban differences in populist vote shares, the actual geographies of the electoral outcomes are more complex than that. Neoliberal globalisation and the ensuing deindustrialisation and immigration have had geographically varying impacts that need to be considered in macro studies of electoral outcomes to examine the contribution of local and regional context that may be more important than the national context to influence voting decisions. The neglect of the subnational scale in explaining recent electoral outcomes is even more surprising if we consider the enormous geographic variation in the populist vote share observed in those three electoral decisions (see Figure 1a–c). The populist vote shares ranged from 13 to 85% in Austrian municipalities, from 21.3 to 75.6% in British Local Authorities and from 4.1 to 95.3% in US counties. In the following, we discuss the data and methods we used to explain those differences. Figure 1. View large Download slide View large Download slide View large Download slide (a) The geographical distribution of the share of the vote for Norbert Hofer in the second run-off to the Austrian presidential elections, 2016 (municipalities). (b) The geographical distribution of the share of votes for Donald Trump in the 2016 US presidential elections (US counties). (c) The geographical distribution of the Brexit vote share, 2016 (local authorities). Figure 1. View large Download slide View large Download slide View large Download slide (a) The geographical distribution of the share of the vote for Norbert Hofer in the second run-off to the Austrian presidential elections, 2016 (municipalities). (b) The geographical distribution of the share of votes for Donald Trump in the 2016 US presidential elections (US counties). (c) The geographical distribution of the Brexit vote share, 2016 (local authorities). Data and method Data for the analysis are obtained from a variety of sources. For Austria, data on socio-demographic factors come from the Austrian Census and were obtained from Statistik Austria (www.statistik-austria.at), while data on elections were taken from a recently established database on Austrian election results (www.wahldatenbank.at). For the UK, the main data source was the Office of National Statistics (www.ons.gov.uk), while the 1980 manufacturing data were retrieved from the UK data service (http://casweb.ukdataservice.ac.uk/) and election data came from the electoral commission (www.electoralcommission.org.uk). For the US, socio-economic data were compiled from the 5-year pooled American Community Surveys, 1980 manufacturing data from the County Business Patterns from the US Census Bureau (www.census.gov), unemployment figures from the Bureau of Labour Statistics (www.bls.gov), the 2016 election results from (townhall.com/election/2016/president) and the 2012 election results from ‘The Guardian’ (www.theguardian.com/news/datablog/2012/nov/07/us-2012-election-county-results-download#data). In all three cases, local authority, county and municipality border changes had to be taken into account and were processed to guarantee that data were available for consistent geographies over time. Our dependent variable was the share of the Brexit vote in 348 local authorities in England and Wales, the share of the vote for the presidential candidate of the Austrian Freedom Party in 2122 municipalities and the share of the Trump vote in 3108 counties of the continental US. To account for the spatial differences in the populist vote, we develop the following, exploratory models:  Vote%=α0+β1IMM+β2STRUCT+β3CYCL+β4ECON+β5X+β6D+ε (1) where Vote% is the share of the populist vote, α0 is a constant, IMM represents as set of variables capturing immigration, STRUCT is a set of variables representing long-term structural economic conditions, CYCL are variables that account for the impact of cyclical economic conditions, the impact of the Great Recession on unemployment and recovery from it, ECON is a vector referring to current economic conditions, X represents socio-demographic controls such as age and gender, and population size D is a set of regional dummy variables controlling for broad regional differences in the populist vote and ε is an error term that may be heteroskedastic of unknown form. While the focus of our models is on economic context, political discourse prior to the three electoral decisions was strongly influenced by various forms of anti-immigrant statements. At the national-scale immigrants are easy scapegoats for failed economic policies, but at the subnational scale, it will depend on whether immigrants settle in regions that benefit or suffer from global economic integration and on the pace of the increase of migrant shares. Another question is whether the origin of immigrants plays a role in explaining the populist vote shares. While UKIP was concerned primarily with the perceived negative effects on public services because of migration from Eastern European EU accession countries (which was also interpreted as EU imposed loss of control of national borders), Austria appeared more worried about a perceived ‘refugee crisis’, while in the US, immigration from Latin America, in particular Mexico, became a key issue in the elections. Hence, we use different variables to represent the impact of immigration, IMM. For the UK, rather than using the share of foreign-born people, we distinguish between the share of people from EU15 countries (joined the EU prior to 2003), eight Eastern European EU accession countries, and the rate of increase in the share of residents from EU accession countries. In the US, we consider the share of the non-white Hispanic population and the increase in the share of the Hispanic population. Although most people of Hispanic origin are not immigrants, they form part of the visible minority that was singled out by Trump, so that we considered the share of this group as more important than the share of foreigners to explain the Trump vote shares. Empirically, the percent of foreign-born and Hispanics are highly correlated. In Austria, because of the small absolute numbers of visible ethnic minorities and refugees, we had little choice but choose the share of foreign-born and the rate of increase of foreign-born as our independent variables representing immigration. We are agnostic about the direction of the influence of the share of immigration variables because it is unclear what the actual effect of increased contact with immigrants is on the perception of them in the voting population. On the one hand, the contact hypothesis states that increasing contact with foreigners reduces the fear and misperceptions about them and, hence, would make it less likely to vote for an anti-immigration policy, but on the other hand, a higher share of foreigners may be considered as competition for jobs and pressure on public services, which should increase the populist right-wing vote share. Because perceived and actual pressure on public services will also depend on the increase rather than level of immigration, we include the rate of increase in immigration as well. For the UK, we look at the increase in immigrants from eight Eastern European EU accession countries, in the US, the increase in the share of people of Hispanic origin and in Austria, simply the rate of change of the foreign-born population. We expect that the rate of increase in immigration is positively related to the populist right-wing vote share. Our set of economic variables corresponds to the long-term economic changes identified as important by the ‘modernisation theorists’, the cyclical effects identified by the ‘economic crisis’ theorists, and current economic conditions that, at the regional level, are the outcome of those long and medium to short-term changes. Our structural economic variables, STRUCT, include the share of the manufacturing sectors in 1980 in the US, 1981 in the UK and 1991 in Austria,6 as well as the rate of change in manufacturing employment between the base year and the most recent year data were available. Choosing manufacturing shares from the 1980s rather than current manufacturing shares allows us to identify old industrial regions. The change in manufacturing employment was included to examine whether the pace of industrial restricting has an influence on voting behaviour. Following Eribon (2009) and Kriesi et al. (2006), who propose that a shrinking traditional working class and its political abandonment by centrist and socialist parties result in resentment and shift of traditional working class votes to the right, we would expect a positive relationship between the original manufacturing share and the populist vote share. We also expect a positive relationship between the rate of change in manufacturing employment and the populist vote because a slower decline indicates a slow pace of restructuring and in turn, fewer local job opportunities for previous manufacturing workers. To depict the impact of economic crisis, CYCL, in our case the Great Recession, on the populist vote share, we use the rise of local unemployment rates until the peak of the recession and the reduction of unemployment after the recession as proxies. We would expect that regions where the recession had a stronger negative impact on employment (a faster rise in unemployment during the recession) and where labour markets recovered slowly from the recession (a slower rate of decline in unemployment rates) exhibit a higher popular vote share. The current economic situation, ECON, is represented through the current unemployment rate. Median household income is highly negatively correlated with the unemployment rate and could not be included. Ideally, we would also have liked to examine the impact of rising inequality on voting behaviour, but unfortunately, information on that was not available for the UK. Instead, we were able to examine the local impact of austerity programs in the UK (data for the US and Austria were not available). In general, we would expect higher unemployment rates and higher losses in benefits per person to result in more dissatisfaction with the status quo and therefore be positively related to populist vote shares. However, if it is true that voters trust established parties and candidates more on economic issues, a higher unemployment rate could also result in a lower share of the populist vote (Ivarsflaten, 2005). Finally, we also include age and gender controls as well as regional population size. We also included education and the percent of White British (for the UK) and White (for the US) in the basic model. However, the problem is the high correlation between the share of the highly educated and our economic and immigration target variables. Hence, we could not control for higher education and white ethnicity in our final models. To account for broad regional trends in voting patterns, we included a Wales and London dummy in the UK case and region dummies for eight of the nine census regions in the US case (we omitted a dummy for the mid-Atlantic states). All three elections built on earlier gains of populist agendas. In the UK and in Austria, the share of the populist right-wing agendas shifted dramatically upwards from previous gains of the FPÖ in Austria and UKIP in the UK. On the other hand, Trump, as a populist but Republican candidate, could rely on the Republican core vote and attack floating voters in specific states through populist messages. In a second step, we therefore explore how well our economic variables do to explain gains or losses in the populist vote through the following model:  Votediff=α0+β1IMM+β2STRUCT+β3CYCL+β4ECON+β5X+β6D+ε (2) where Votediff refers to the difference in regional vote shares between previous elections and the most recent elections. In the UK, we subtracted the share of the UKIP vote of the European 2014 Parliamentary Elections from the Brexit vote, in the US, we subtracted the share of the Romney vote in the 2012 US presidential elections from the 2016 Trump share, and in Austria, we subtracted the share of the FPÖ vote of the last general election from the share of the Hofer vote. We made a minor adjustment to model (1) to highlight recent economic conditions by replacing the historic manufacturing share and the rate of change with the current manufacturing share. This was motivated also by Trump’s success of targeting current manufacturing workers in the old industrial heartland of the US. Because data for the US are based on samples, we used population weights to minimise measurement error in the US case. The exact specification of our variables is summarised in Appendix A. Empirical results Before discussing our results, we offer a table with simple correlation coefficients between regional socio-demographic and economic characteristics identifying the core voter potential of populist right parties for the three countries. While populist parties tend to rest on relatively fluid and poorly defined social bases, Arzheimer (2011) argues that the electorate of the populist right in different countries nevertheless does consist of a clearly defined social core. They tend to be overwhelmingly male, poorly educated, ‘lower’ social classes (petty-bourgeoisie of artisans, shopkeepers, farmers and other self-employed are increasingly supplemented with non-traditional workers, members of the lower middle classes and the unemployed) and younger voters. Table 1 shows how the spatial distribution of those ‘core characteristics’ correlate with the populist vote shares in our three country cases. Because occupation is highly correlated with education and because this variable is not readily available on a consistent basis for the three countries, we omit occupation. We also add ‘ethnicity’ and population size of regions, as the strong nativist tone in the three elections should resonate more with the ‘native’, ‘pure’ population which is defined as white in the UK and US case. In Austria, we simply look at the share of foreign-born, as the share of immigrants that would fall under ‘non-white’ would be negligible. As the campaigns were also run on a traditional/cosmopolitan cleavage, we include regional size as we believe that large urban areas tend to be more cosmopolitan in outlook, while rural areas tend to be more inward looking and traditional. Table 1. Simple correlation coefficients between “core populist right voter characteristics” and populist vote shares.   Hofer %  Brexit %  Trump %  Young  −0.12  −0.30  −0.21  Old  +0.14  +0.41  +0.41  Male  +0.15  −0.20  +0.30  High education  −0.56  −0.89  −0.57  Medium education  n.a.  +0.78  +0.64  Low/no education  +0.30  +0.76  −0.07  Ethnicity  n.a.  +0.54  +0.67  Population size  −0.14  −0.18  −0.69  Number of regions  2122  348  3108    Hofer %  Brexit %  Trump %  Young  −0.12  −0.30  −0.21  Old  +0.14  +0.41  +0.41  Male  +0.15  −0.20  +0.30  High education  −0.56  −0.89  −0.57  Medium education  n.a.  +0.78  +0.64  Low/no education  +0.30  +0.76  −0.07  Ethnicity  n.a.  +0.54  +0.67  Population size  −0.14  −0.18  −0.69  Number of regions  2122  348  3108  Young: UK: % 18–24 year olds; US: % 20–24 year olds; AT: % 20–24 year olds; old: % >65 year olds; male: % of male population; high education: % with college degree or higher; medium education: UK: % with some education but no college degree; US: % with high school degree; low education: UK: % with no qualifications; US: % with less than high school degree; AT: % with apprenticeship or no formal qualification; ethnicity: UK: % white British; US: % white; population size: logarithm of total population. In the US, observations are weighted by population size. View Large The most striking result from Table 1 is the high correlation of the education variables with the Brexit vote and, to a lesser extent, the Trump and Hofer votes. In the UK, the share of college graduates correlates highly negatively with the Brexit vote. The negative coefficients are smaller for the Trump and Hofer votes. In the UK and US, the share of those with medium education levels is highly positively correlated with the Trump and Brexit vote shares, and the share of those with no or low qualifications is positively correlated with the populist vote in the UK and Austrian cases but insignificant in the US case. In all cases, the share of the young is negatively correlated with the populist vote share, while the share of the old is positively correlated. Correlations for the percentage of males are small and have different signs, which can be explained by the small spatial variation of this variable. The share of white British and Whites is strongly correlated with the Brexit and Trump vote share. In the US, the size of a county exhibits the highest correlation coefficient with the populist vote indicating that the ‘traditional/cosmopolitan’ distinction is particularly relevant there. Based on these simple correlations we could build a model around the relationship between education and the populist vote share. This would certainly yield a good model result, but what would it actually tell us? As the highly educated tend to occupy managerial and professional jobs that yield high salaries and whose jobs are less threatened by globalisation and immigration it seems plausible that they are less likely to vote for change or anti-globalisation or -immigration politics and policies. It could also be the case that the highly educated tend to be more liberal and hence, are less likely to vote for the authoritarian policies of the populist right. Or, at the level of regions, does it mean that their skills allow them to move to urban centres that tended to benefit from economic integration at the expense of rural areas and old industrial regions, and that these economic opportunities rather than individual attributes make them vote against populist agendas? In this case, the share of the highly educated correlates with a number of other regional characteristics that we would expect to influence voting outcomes. Indeed, if we look more carefully at the relationships between the share of the highly educated and our target variables, then the share of the highly educated is highly correlated with the share of foreign-born (and in the US with city size), but highly negatively correlated with unemployment rates or manufacturing employment shares in regions. Similarly, the share of white British and Whites is strongly negatively correlated with immigration and ethnic minority variables. In an ideal world, we would combine individual-level data with regional contextual data to isolate different processes. But as we do not have access to detailed individual-level data, we choose to model more explicitly the impact of regional economic context and immigration and explore how much those variables account for regional variation in the populist vote shares. Table 2 presents the results of a simple OLS regression, including robust standard errors and beta values. The dependent variable is the percentage of the populist vote. Parameter estimates for control variables age and gender as well as region dummies and the constants are omitted to allow a better focus on the target variables. The first observation is that the results are fairly robust across the three electoral decisions. The parameter estimates for the level of immigration (the share of foreign-born in Austria, the share of EU15 born in the UK, the share of Hispanics in the US) are negative and significant. Their relative effects are also high in all three countries. While the level of Eastern European accession country-born is insignificant in the UK, the increase in immigration (increase in foreign-born, increase in Eastern European accession state citizens, increase in Hispanics) is significant and positive in all three electoral outcomes, although the effect is substantially lower than for the negative impact of the share of immigrants. This result is consistent with Goodwin and Heath (2016), who identified the immigration accelerator effect for the UK. Table 2. Parameter estimates for Model (1), with robust standard errors in parentheses and beta values in brackets. Dependent variables    % Hofer  % Brexit  % Brexit  % Trump  Immigration  Level  −0.454**  −2.327**  −2.501**  −0.214**      (0.044)  (0.787)  (0.727)  (0.035)      [−0.287]  [−0.356]  [−0.382]  [−0.208]    EU accession countries, level    −0.328  −0.621          (0.337)  (0.320)          [−0.053]  [−0.100]      Change  0.002*  0.176**  0.187**  2.597**      (0.002)  (0.055)  (0.055)  (0.518)      [0.044]  [0.127]  [0.134]  [0.135]  Structural  Manufacturing %, base year  0.178**  0.269**  0.255**  0.343**      (0.023)  (0.055)  (0.054)  (0.037)      [0.190]  [0.239]  [0.227]  [0.261]    Manufacturing change  0.006  0.213**  0.185**  0.018**      (0.019)  (0.039)  (0.037)  (0.003)      [0.010]  [0.271]  [0.235]  [0.145]  Cyclical/recession  Impact  −0.011**  0.103**  0.110**  0.070**      (0.006)  (0.018)  (0.016)  (0.012)      [−0.066]  [0.237]  [0.253]  [0.202]    Recovery  0.015**  0.064  0.148**  0.131**      (0.007)  (0.046)  (0.036)  (0.049)      [0.060]  [0.052]  [0.120]  [0.069]  Current  Unemployment rate  0.545**  3.554**    −0.118      (0.120)  (0.501)    (0.343)      [0.171]  [0.317]    [−0.010]    Benefit loss per person      0.027**            (0.003)            [0.340]    Size  Log of population size  −0.009**  −2.581**  −2.802**  −4.093**      (0.003)  (0.664)  (0.612)  (0.460)      [−0.087]  [−0.147]  [−0.159]  [−0.383]  Regional dummies    No  Yes  Yes  Yes  Population weights    No  No  No  Yes  R2    0.19  0.72  0.74  0.69  Dependent variables    % Hofer  % Brexit  % Brexit  % Trump  Immigration  Level  −0.454**  −2.327**  −2.501**  −0.214**      (0.044)  (0.787)  (0.727)  (0.035)      [−0.287]  [−0.356]  [−0.382]  [−0.208]    EU accession countries, level    −0.328  −0.621          (0.337)  (0.320)          [−0.053]  [−0.100]      Change  0.002*  0.176**  0.187**  2.597**      (0.002)  (0.055)  (0.055)  (0.518)      [0.044]  [0.127]  [0.134]  [0.135]  Structural  Manufacturing %, base year  0.178**  0.269**  0.255**  0.343**      (0.023)  (0.055)  (0.054)  (0.037)      [0.190]  [0.239]  [0.227]  [0.261]    Manufacturing change  0.006  0.213**  0.185**  0.018**      (0.019)  (0.039)  (0.037)  (0.003)      [0.010]  [0.271]  [0.235]  [0.145]  Cyclical/recession  Impact  −0.011**  0.103**  0.110**  0.070**      (0.006)  (0.018)  (0.016)  (0.012)      [−0.066]  [0.237]  [0.253]  [0.202]    Recovery  0.015**  0.064  0.148**  0.131**      (0.007)  (0.046)  (0.036)  (0.049)      [0.060]  [0.052]  [0.120]  [0.069]  Current  Unemployment rate  0.545**  3.554**    −0.118      (0.120)  (0.501)    (0.343)      [0.171]  [0.317]    [−0.010]    Benefit loss per person      0.027**            (0.003)            [0.340]    Size  Log of population size  −0.009**  −2.581**  −2.802**  −4.093**      (0.003)  (0.664)  (0.612)  (0.460)      [−0.087]  [−0.147]  [−0.159]  [−0.383]  Regional dummies    No  Yes  Yes  Yes  Population weights    No  No  No  Yes  R2    0.19  0.72  0.74  0.69  Immigration level refers to percent of EU15 in UK, percent Hispanics in the US and percent foreign-born in Austria; change refers to rate of change in population from Eastern European EU accession countries in UK, difference in share of Hispanics in the US and rate of change of foreign-born in Austria; impact of recession refers to increase in unemployment rates prior to recession and recover refers to reduction in unemployment rate since the recession. ** (*) significant at 0.01 (0.05) level. View Large Old industrial areas, and especially those that have not reduced their manufacturing shares quickly enough, exhibit higher shares of the populist vote. Again, this result is consistent across all three countries (see also Shafique, 2016 for the UK). Moving on to the impact of the recession, we observe a positive relationship between the increase in unemployment rate going into the recession in the US and UK and, to a lesser extent, a positive relationship between slow recovery (unemployment rates decline at a slower pace) and the populist vote shares. In Austria, somewhat surprisingly, the increase in the local unemployment rates during the recession is associated negatively and a slow recovery is barely related to the Hofer share. This may have to do with the different evolution of Austrian unemployment rates. There, unemployment rates increased steadily until 2009, then dropped briefly until 2011, at which point they began to rise again. In the UK and US, the impact of the recession on the labour markets and its recovery from it were much more severe, and unemployment rates followed a clearer path of increase and decline prior and after the peak of the recession. Overall, the relative effect of the recession on the populist vote appears slightly less important than the historical importance of the manufacturing sector in a region. The final set of economic variables is the current unemployment rate and the impact of austerity measures (for the UK). In Austria and the UK, local, current unemployment rates have a big relative effect on populist vote shares, and this after controlling for the impact of the Great Recession. In the US, the current unemployment rate has no significant impact on the Trump vote share. This appears surprising, but may be explained by the fact that the core Republican voter is less concerned with unemployment and more with cultural and social policies. If this were the case, then unemployment should become important when explaining gains in the Trump over Romney vote. Because in the UK austerity measures were severe and varied substantially across localities, we also offer the results of a version that replaced the current unemployment rate with the average annual loss in benefits per person in each region.7 Apart from the share of EU15 born, the loss of benefits has the largest relative effect in the UK model. And finally, to account for a ‘cosmopolitan’ effect of cities that goes beyond the share of foreigners, we also included the population size as independent variable. This variable is significant and negative and particularly important to explain the Trump vote share in the US (this is the case for the weighted and unweighted models). While we do not dwell on the regional effects in the US elections, it is worth mentioning that the indicator variable for Wales was strongly negative and that for London positive. While the estimate for Wales does not come as a surprise, the positive value for London does. It suggests that given the low manufacturing share, lower unemployment rates and high share of immigrants, the Brexit voter share was, on average, 6.8% higher in London boroughs than we would expect given those conditions. The model fits for the US and the UK are reasonably high. For the Austrian case, the model fit is relatively poor. A population-weighted model would have pushed up the R2, but since we rely on census and not sample data, we did not consider that appropriate for the Austrian case. We now examine the enormous increase in the populist vote from previous elections (in the case of Austria and the UK) and to get a better understanding where and why Trump added to the Republican core vote. Table 3 provides the results of this analysis. Our dependent variables are now differences in the Hofer, Brexit and Trump vote shares from the FPÖ, UKIP and Romney vote shares in the last Austrian general elections, the 2014 EU parliamentary elections and the 2012 US presidential elections. Table 3. Parameter estimates for Model (2)   Austria  UK  US  Midwest  Immigration  Level  −0.487**  −1.425**  0.004  −0.148      (0.044)  (0.388)  (0.013)  (0.076)      [−0.326]  [−0.419]  [0.011]  [−0.185]    EU2    0.426            (0.263)            [0.132]        Change  0.003  −0.036  0.004  −0.266      (0.002)  (0.038)  (0.013)  (0.470)      [0.044]  [−0.050]  [−0.038]  [−0.027]  Structural  Manufacturing share, current  0.024  0.497**  0.097**  0.244**      (0.027)  (0.080)  (0.028)  (0.030)      [0.019]  [0.388]  [0.090]  [0.244]  Cyclical/recession  Impact  −0.010*  −0.008  0.003  −0.052**      (0.004)  (0.012)  (0.006)  (0.009)      [−0.060]  [−0.035]  [0.028]  [−0.229]    Recovery  0.021**  −0.080**  −0.026  −0.031      (0.006)  (0.030)  (0.019)  (0.022)      [0.088]  [−0.124]  [−0.041]  [−0.051]  Current  Unemployment rate  0.438**  2.578**  1.068**  1.869**      (0.111)  (0.335)  (0.190)  (0.147)      [0.146]  [0.443]  [0.285]  [0.397]  Size  Log of population size  −0.018**  −0.542  −1.132**  −1.394**      (0.003)  (0.436)  (0.133)  (0.199)      [−0.175]  [−0.059]  [−0.320]  [−0.398]  Regional dummies    No  Yes  Yes  No  Population weights    No  No  Yes  Yes  R2    0.30  0.61  0.60  0.74    Austria  UK  US  Midwest  Immigration  Level  −0.487**  −1.425**  0.004  −0.148      (0.044)  (0.388)  (0.013)  (0.076)      [−0.326]  [−0.419]  [0.011]  [−0.185]    EU2    0.426            (0.263)            [0.132]        Change  0.003  −0.036  0.004  −0.266      (0.002)  (0.038)  (0.013)  (0.470)      [0.044]  [−0.050]  [−0.038]  [−0.027]  Structural  Manufacturing share, current  0.024  0.497**  0.097**  0.244**      (0.027)  (0.080)  (0.028)  (0.030)      [0.019]  [0.388]  [0.090]  [0.244]  Cyclical/recession  Impact  −0.010*  −0.008  0.003  −0.052**      (0.004)  (0.012)  (0.006)  (0.009)      [−0.060]  [−0.035]  [0.028]  [−0.229]    Recovery  0.021**  −0.080**  −0.026  −0.031      (0.006)  (0.030)  (0.019)  (0.022)      [0.088]  [−0.124]  [−0.041]  [−0.051]  Current  Unemployment rate  0.438**  2.578**  1.068**  1.869**      (0.111)  (0.335)  (0.190)  (0.147)      [0.146]  [0.443]  [0.285]  [0.397]  Size  Log of population size  −0.018**  −0.542  −1.132**  −1.394**      (0.003)  (0.436)  (0.133)  (0.199)      [−0.175]  [−0.059]  [−0.320]  [−0.398]  Regional dummies    No  Yes  Yes  No  Population weights    No  No  Yes  Yes  R2    0.30  0.61  0.60  0.74  Dependent variables are the differences in vote shares from previous elections (UKIP vote for EU parliamentary elections in the UK, percent Romney vote in the US; FPÖ vote in Austria). Robust standard errors are in parenthesis and beta values in brackets. Immigration level refers to percent of EU15 in UK, percent Hispanics in the US and percent foreign-born in Austria; change refers to rate of change in population from Eastern European EU accession countries in UK, difference in share of Hispanics in the US and rate of change of foreign-born in Austria; impact of recession refers to increase in unemployment rates prior to recession and recover refers to reduction in unemployment rate since the recession. ** (*) significant at 0.01 (0.05) level. View Large Table 3 highlights a number of issues. The share of immigrants is negatively related to the populist vote share gains in Austria and the UK, but has no impact on Trump’s vote gain. The increase in immigration had no further impact on increasing the populist vote shares. Those for whom rising immigration was an issue voted populist prior to the recent electoral decisions, it appears. The share of current manufacturing employment is important in the UK (Shafique, 2016) and the US, in particular in the Midwest. In Austria, the impact of the manufacturing share is insignificant. The current unemployment rate is important to explain the populist vote share increase in all three elections, in particular the UK and the American Midwest. In other words, while the unemployment rate is not important to explain the basic Republican vote share, it is important to explain the gains Trump made over Romney. Population size is strongly negatively related to populist vote gains in Austria and the US but not in the UK. The impact of the recession on labour markets appears to have had little effect on increasing the populist vote share. Current unemployment rates are more important. Overall, the model performs best to explain the Trump gains in the Midwest, where his team targeted areas with sizeable manufacturing bases and where the message of protecting American interests and industries fell on fertile ground. Conclusion Drawing inspiration from macro-level studies on the role of modernisation processes and economic crisis on the rise of populist right parties, candidates and agendas, we examined the impact of subnational economic context on the regional share of the Hofer, Brexit and Trump votes. We believe that more attention to the subnational scale is important to obtain a tighter fit between the theoretical arguments linking the effects of socio-economic and cultural change to individual voting behaviour because modernisation processes in form of globalisation, deindustrialisation and migration affect regions and in turn, the environment in which individuals make political choices, differently. Aggregate studies at the national scale gloss over those differences. Our results demonstrate that economic variables perform reasonably well to explain subnational variation in the Brexit and Trump votes, but less so to explain regional variation in the Hofer vote. Regional shares in the populist right-wing vote tend to be higher in old industrial regions, are more pronounced in those that restructure more slowly, tend to be higher in areas with high current unemployment rates, those areas whose labour markets were affected more severely by the Great Recession, and they were lower in larger regions and those that were characterised by higher shares of foreign-born. Importantly though, regions where immigrant shares rose faster tended to exhibit higher shares of the populist right-wing vote. This may indicate that rapidly rising immigrant shares increase the demand for local services. Or, it may mean that austerity measures reduce the supply of local services and benefits that will be felt the most in regions with increased demand for these services. For the UK, we could show that faster increases in immigrants from EU accession countries and higher losses in benefits had a positive effect on the regional Brexit vote share. While these first-cut results are promising, the interaction between the supply of and demand for public services and welfare benefits and their effects on the populist right-wing vote share in different regions and countries needs further, detailed investigation. When switching the focus to increases in vote shares, we attempted to examine the impact on floating rather than core voters of populist candidates and parties. Here, current economic conditions seemed more important than the cyclical effects or increases in immigration. In particular, the current unemployment rate, the level of migration and, for the US, a high share in manufacturing employment was important to explain those vote increases. For the US, the share of manufacturing was a stronger explanatory factor in the Midwest than the country as a whole, which suggests that the supply of a populist agenda targeting disenfranchised manufacturing workers in the old industrial heartland was important for the outcome of the election. Not too surprisingly, there are a number of problems that affect almost all cross-sectional analyses. Without access to individual-level data, we are unable to establish causal links between our economic variables and individual voting behaviour. However, we also believe that the results of this exploratory analysis are encouraging and invite further work in the following directions: First, the individual country case studies need to be developed in greater detail drawing, for instance, on historical voting records and taking into account migration patterns to separate cause from effect: is it that individuals become more open to change as a result of their exposure to diversity and hence, are less likely to vote for populist right ideas, or are those that are more open to new ideas more likely to leave structurally declining areas and move to more diverse cities where they continue to vote against the populist right? Second and related, noting that parameter estimates are lower but in line with the US and UK case, what may explain the poor model fit for Austria? One explanation may be that the Austrian manufacturing and tourism sectors benefitted from joining the EU, that parts of the old nationalised metals and steel industries have been restructured successfully and that the small- and medium-sized enterprise sector has been thriving, at least prior to the recession. Because of the relatively high standard of living, even among the unskilled and less educated, and the lower levels of spatial and social inequality, economic explanations may be less successful in explaining the populist vote in Austria. One of the key groups of FPÖ voters are the ‘welfare chauvinists’ (Fallend, 2013), and it may be the fear of losing those welfare benefits to immigrants and refugees rather than the actual, local economic context that motivated Austrians to vote for Hofer. However, these hypotheses need to be explored and corroborated through further analysis. Third, as one of the key changes since the 1980s is increasing inequality and social polarisation, it would be important to examine whether inequality has an effect on the populist vote shares. Fourth, future work needs to look more carefully at interactions between different variables. For instance, what is the impact of the level and increase of immigration on the populist right-wing vote share in high and low unemployment areas, in fast and slow growing regions, in old industrial regions and new industrial spaces? Or, what is the impact of unemployment or low education levels on the populist vote shares in high growth and low growth regions? What is the impact of welfare losses in high and low immigration areas? Again, those questions are easier carried out for individual countries first. Acknowledgements We thank Maria Reyero for her assistance in compiling data for the US and Anna Stelzer for her assistance in compiling the data set for Austria. Data collection for the Austrian case was supported by the WU Small Scale Projects scheme, grant no. 11000416. We also would like to thank the editorial team for their support in turning around the script in record time. References Arzheimer, K. ( 2011) Electoral sociology: who votes for the extreme right and why – and when? In U. Backes and P. Moreau (eds) The Extreme Right in Europe: Current Trends and Perspectives , pp. 35– 50. Göttingen: Vandenhoeck and Ruprecht. Google Scholar CrossRef Search ADS   Autor, D., Dorn, D., Hanson, G. and Majlesi, G. ( 2016) A Note on the Effect on Rising Trade Exposure on the 2016 Presidential Elections . Available online at: https://gps.ucsd.edu/_files/faculty/hanson/hanson_research_TrumpVote-032017.pdf [Accessed 4 April 2017]. Becker, S. O., Fetzer, T. and Novy, D. ( 2016) Who Voted For Brexit? A Comprehensive District-Level Analysis . Coventry: Centre for Competitive Advantage in the Global Economy, Department of Economics, The University of Warwick. Working Paper Series, October 2016, No. 305. Betz, H. G. ( 1994) Radical Right-Wing Populism in Western Europe . New York, NY: St. Martin’s Press. Google Scholar CrossRef Search ADS   Blyth, M. ( 2016) Global Trumpism: Why Trump’s Victory Was Thirty Years in the Making and Why It Won’t Stop Here . Foreign Affairs, November 2016. Available online at: https://www.foreignaffairs.com/articles/2016-11-15/global-trumpism. Blyth, M. and Matthijs, M. ( 2017) Black swans, lame ducks, and the mystery of IPE’s missing macroeconomy, Review of International Political Economy , 24: 203– 231. Google Scholar CrossRef Search ADS   Eribon, D. ( 2009) Die Rückkehr nach Reims . Frankfurt: Suhrkamp. Esping-Anderson, G. ( 1990) Three Worlds of Welfare Capitalism . Princeton, NJ: Princeton University Press. Evans, J. A. J. ( 2005) The dynamics of social change in radical right-wing populist party support, Comparative European Politics , 3: 76– 101. Google Scholar CrossRef Search ADS   Evans, G. and Mellon, J. ( 2016) Working class votes and conservative losses: solving the UKIP puzzle, Parliamentary Affairs , 69: 464– 479. Google Scholar CrossRef Search ADS   Fallend, F. ( 2013) Populism in government: the case of Austria (2000–2007). In C. Mudde and R. Kaltwasser (eds) Populism in Europe and the Americas , pp. 113– 135. Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Ford, R. and Goodwin, M. ( 2017) A nation divided, Journal of Democracy , 28: 17– 30. Google Scholar CrossRef Search ADS   Golder, M. ( 2003) Explaining variation in the electoral success of extreme right parties in Western Europe, Comparative Political Studies , 36: 432– 466. Google Scholar CrossRef Search ADS   Golder, M. ( 2016) Far right parties in Europe, Annual Review of Political Science , 19: 477– 497. Google Scholar CrossRef Search ADS   Goodwin, M. J. and Heath, O. ( 2016) The 2016 referendum, Brexit and the left behind: an aggregate-level analysis of the result, The Political Quarterly , 87: 323– 332. Google Scholar CrossRef Search ADS   Harris, R. and Charlton, M. ( 2016) Voting out of the European Union: exploring the geography of leave, Environment and Planning A , 48: 2116– 2128. Google Scholar CrossRef Search ADS   Harvey, D. ( 2005) A Brief History of Neoliberalism . Oxford: Oxford University Press. Hobolt, S. B. ( 2016) The Brexit vote: a divided nation, a divided continent, Journal of European Public Policy , 23: 1259– 1277. Google Scholar CrossRef Search ADS   Inglehart, R. F. ( 1990) Cultural Shift in Advanced Industrial Society . Princeton, NJ: Princeton University Press. Inglehart, R. F. ( 1997) Modernization and Postmodernization: Cultural, Economic and Political Change in 43 Societies . Princeton, NJ: Princeton University Press. Inglehart, R. F. and Norris, P. ( 2016) Trump, Brexit, and the Rise of Populism: Economic Have-nots and Cultural Backlash . Faculty Research Working Paper Series. RWP16-026. August. Cambridge, MA: Harvard Kennedy School. Ivarsflaten, E. ( 2005) The vulnerable populist right parties: no economic realignment fuelling their electoral success, European Journal of Political Research , 44: 465– 492. Google Scholar CrossRef Search ADS   Johnson, P. ( 2015) Older Industrial Britain Being Left Behind. Available online at: https://newstartmag.co.uk/articles/older-industrial-britain-being-left-behind-in-economic-upturn/ [Accessed 3 September 2017]. Kriesi, H. P., Grande, E., Lachat, R., Dolezal, A., Bornschi, S. and Frey, T. ( 2006) Globalization and the transformation of the national political space: six European countries compares, European Journal of Political Research , 45: 921– 956. Google Scholar CrossRef Search ADS   Kriesi, H. P., Koopmans, R., Duyvendak, J. W. and Giugni, M. G. ( 1995) New Social Movements in Western Europe: A Comparative Analysis . Minneapolis, MN: University of Minnesota Press. Lipset, S. M. ( 1960) Political Man: The Social Basis of Politics . New York, NY: Doubleday. Los, B., McCann, P., Springford, J. and Thissen, M. ( 2017) The mismatch between local voting and the local economic consequences of Brexit, Regional Studies , 51: 786– 799. Google Scholar CrossRef Search ADS   Lubbers, M., Gijsberts, M. and Sheepers, P. ( 2002) Extreme right-wing voting in Western Europe, European Journal of Political Research , 41: 345– 378. Google Scholar CrossRef Search ADS   McCarty, N., Poole, K. T. and Rosenthal, H. ( 2016) Polarized America . Cambridge, MA: The MIT Press. Minkenberg, M. ( 2000) The renewal of the radical right: between modernity and anti-modernity, Government Opposition , 35: 170– 188. Google Scholar CrossRef Search ADS   Mouffe, C. ( 2016) The Populist Moment . Available online at: https://www.opendemocracy.net/democraciaabierta/chantal-mouffe/populist-moment [Accessed 11 January 2017]. Mudde, C. ( 2007) Populist Radical Right Parties in Europe . New York, NY: Cambridge University Press. Google Scholar CrossRef Search ADS   Mudde, C. and Kaltwasser, R. ( 2013) Populism and (liberal) democracy: a framework for analysis. In C. Mudde and R. Kaltwasser (eds) Populism in Europe and the Americas , pp. 1– 26. Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Müller, W. C. ( 2002) Evil or the ‘engine of democracy’? Populism and party competition in Austria. In Y. Meny and Y. Surel (eds) Democracies and the Populist Challenge , pp. 155– 175. Basingstoke: Palgrave. Google Scholar CrossRef Search ADS   Müller, J.-W. ( 2016) Was ist Populismus?  Berlin: Suhrkamp. Norris, P. and Inglehart, R. J. ( 2009) Cosmopolitan Communications . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Oliver, J. E. and Rahn, W. M. ( 2016) Rise of the Trumpenvolk: populism in the 2016 Election, The Annals of the American Academy of Political and Social Science , 667: 189– 206. Google Scholar CrossRef Search ADS   Piketty, T. ( 2014) Capital  in the 21st Century. Cambridge, MA: Belknap Press. Pool, K. and Rosenthal, H. ( 2009) Ideology and Congress . New Brunswick and London: Transaction Publishers. Shafique, A. ( 2016) BREXIT Was Driven by Places ‘Left Behind’. Available online at: https://www.thersa.org/discover/publications-and-articles/rsa-blogs/2016/08/brexit-was-driven-by-places-left-behind [Accessed 2 September 2017]. Sheppard, E. ( 2015) The Limits to Globalization . Oxford: Oxford University Press. Singh, R. ( 2017) ‘I, the people’: a deflationary interpretation of populism, Trump and the United States constitution, Economy and Society , 46: 20– 42. Google Scholar CrossRef Search ADS   Taggart, P. ( 2000) Populism . Buckingham: Open University Press. Tyson, A. and Maniam, S. ( 2016) Behind Trump’s Victory: Divisions by Race, Gender, Education . Pew Research Center. Available online at: http://www.pewresearch.org/fact-tank/2016/11/09/behind-trumps-victory-divisions-by-race-gender-education/ [Accessed 10 April 2017]. Endnotes 1 Although some may argue that there are also populist left-wing parties, in the North American and European context, they do not tend to pursue a strong nativist agenda and are not authoritarian. Also, in our case studies, we focus on three populist right-wing candidates and agendas. 2 Inglehart and Norris (2016) show that the average vote share of Western European populist parties in European parliamentary and national elections more than doubled since the 1960s, from 5.1 to 13.2%, at the expense of the centre parties. During the first round of the Austrian presidential elections, the two centre party candidates achieved a combined vote share of only 22.4% and were both eliminated from the run-off elections. 3 Notice though that Bernie Sanders in the US and Jeremy Corbyn in the UK had some electoral success when attempting to spell out those links, such that a failure by traditional parties to engage in this discourse is in part to blame for the simple “Eliminate Unemployment: Stop Immigration” (Golder, 2003: 438) slogan. 4 Officially, Alexander Van der Bellen ran as independent candidate. However, he was a long-standing leading figure of the Austrian Green Party in the past. 5 In the national legislative elections that took place on 15 October 2017, the FPÖ ended up with 26.0% of the vote and is currently in negotiation to form the next government with the conservative ÖVP whose new leader Sebastian Kurz delivered victory (31.5% of the vote) by adopting many of the anti-immigrant elements of the FPÖ’s party programme. The left-centre SPÖ, with 26.9% the second strongest party, will form the parliamentary opposition. 6 Although we would have preferred data on manufacturing employment for 1980, they were not available at municipality level for Austria. 7 Because of the high correlation between the current unemployment rate and benefit losses, we could not include both variables in the same model. Appendix A     Austria  UK  USA  Immigration  Level  Share of foreign-born, 2015  Share of EU15 born, 2011  Share of Hispanics, 2015    EU2    Share of Eastern European EU accession state citizens, 2011      Change  Rate of change in foreign-born, 2001–2015  Rate of change in EU accession citizens, 2001–2011  Difference in Hispanic share, 2005–2015  Structural  Manufacturing %, base year  Share of manufacturing employment, 1991  Share of manufacturing employment, 1981  Share of manufacturing employment, average 1979–1981    Manufacturing change  Rate of change in manufacturing employment, 1991–2015  Rate of change in manufacturing employment, 1981–2011  Rate of change in manufacturing employment, 1980–2015    Manufacturing share, current  Share of manufacturing employment, 2015  Share of manufacturing employment, 2011  Share of manufacturing employment, 2015  Cyclical/ recession  Impact  Rate of change in unemployment, 2001–2009  Rate of change in unemployment, 2001–2011  Rate of change in unemployment, 2005–2010    Recovery  Rate of change in unemployment, 2009–2015  Rate of change in job seekers allowance claimants, 2001–2016  Rate of change in unemployment, 2010–2015  Current  Unemployment rate  Unemployment rate, 2015  Job claimant rate, 2016  Unemployment rate, 2015    Benefit loss per person    Average annual loss in total benefits since benefit reform in 2014    Size  Log of population size  Log of total population, 2015  Log of total population, 2011  Log of total population, 2015      Austria  UK  USA  Immigration  Level  Share of foreign-born, 2015  Share of EU15 born, 2011  Share of Hispanics, 2015    EU2    Share of Eastern European EU accession state citizens, 2011      Change  Rate of change in foreign-born, 2001–2015  Rate of change in EU accession citizens, 2001–2011  Difference in Hispanic share, 2005–2015  Structural  Manufacturing %, base year  Share of manufacturing employment, 1991  Share of manufacturing employment, 1981  Share of manufacturing employment, average 1979–1981    Manufacturing change  Rate of change in manufacturing employment, 1991–2015  Rate of change in manufacturing employment, 1981–2011  Rate of change in manufacturing employment, 1980–2015    Manufacturing share, current  Share of manufacturing employment, 2015  Share of manufacturing employment, 2011  Share of manufacturing employment, 2015  Cyclical/ recession  Impact  Rate of change in unemployment, 2001–2009  Rate of change in unemployment, 2001–2011  Rate of change in unemployment, 2005–2010    Recovery  Rate of change in unemployment, 2009–2015  Rate of change in job seekers allowance claimants, 2001–2016  Rate of change in unemployment, 2010–2015  Current  Unemployment rate  Unemployment rate, 2015  Job claimant rate, 2016  Unemployment rate, 2015    Benefit loss per person    Average annual loss in total benefits since benefit reform in 2014    Size  Log of population size  Log of total population, 2015  Log of total population, 2011  Log of total population, 2015  View Large © The Author(s) 2018. Published by Oxford University Press on behalf of the Cambridge Political Economy Society. All rights reserved. For permissions, please email: journals.permissions@oup.com

Journal

Cambridge Journal of Regions, Economy and SocietyOxford University Press

Published: Mar 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial