Left-behind versus unequal places: interpersonal inequality, economic decline and the rise of populism in the USA and Europe

Abstract

Economic change over the past 20 years has rendered many individuals and territories vulnerable, leading to greater interpersonal and interterritorial inequality. This rising inequality is seen as a root cause of populism. Yet, there is no comparative evidence as to whether this discontent is the consequence of localised interpersonal inequality or stagnant growth in ‘left-behind’ places. This article assesses the association between levels and changes in local GDP per capita and interpersonal inequality, and the rise of far-right populism in Europe and in the USA. The analysis—conducted at small region level for Europe and county level for the USA—shows that there are both similarities and differences in the factors connected to populist voting on both sides of the Atlantic. In the USA, neither interpersonal inequality nor economic decline can explain populist support on their own. However, these factors gain significance when considered together with the racial composition of the area. Counties with a large share of white population where economic growth has been stagnant and where inequalities have increased supported Donald Trump. Meanwhile, counties with a similar economic trajectory but with a higher share of minorities shunned populism. In Europe, the most significant factor behind the rise of far-right populism is economic decline. This effect is particularly large in areas with a high share of immigration.

1. Introduction

The transition in recent decades from an ‘old’ to a ‘new’ economic and production system has created winners and losers (O’Brien and Leichenko, 2003; Hobolt, 2016; Jensen et al., 2017). Across the developed world, many individuals have lost out to technological change and greater trade integration. The jobs they performed, often requiring easily substitutable skills and low levels of training, have been taken over by computers, robots or outsourced and offshored to low-cost countries (Autor et al., 2016; Jensen et al., 2017; Colantone and Stanig, 2018). Wealth is becoming concentrated in a thinner layer of individuals at the top of the income scale (Dorling, 2019). The increase in social polarisation and the growing divide between the ‘haves’ and the ‘have nots’ is triggering widespread discontent (McCann, 2020) and stoking resentment against the system (Cramer, 2016; Wuthnow, 2019). People left-behind are expressing their discontent by casting their votes for populist parties (e.g., Komlos, 2018; Pástor and Veronesi, 2021; Hopkin, 2020).

In parallel to the rise of interpersonal inequalities, the new economic and production system is affecting territories differently, resulting in left-behind places or places that have experienced long periods of economic decline and stagnation (Pike et al., 2023, 4; Rodríguez-Pose and Ketterer, 2020). In a system in which technological progress and advances in logistics allow the direct connection between producers and customers wherever they are in the world and where geopolitical risks abound, geographically spread value chains are becoming less relevant than before. As non-physical assets become more important than physical ones in the ‘new’ economy (Brun et al., 2019), the proceeds of economic activity increasingly concentrate in the headquarters of large firms, located in a handful of superstar cities (Kemeny and Storper, 2020). Hence, the rise in territorial polarisation linked to trade integration—especially in internally heterogeneous countries (Rodríguez-Pose, 2012; Hirte et al., 2020)—is widening the gap between rich and dynamic areas and poor and/or declining regions, a trend that has picked up pace in recent decades (Rosés and Wolf, 2018). The fact that many of the ‘losing’ areas are perceived as having ‘no future’—increasingly becoming ‘places that don’t matter’ (Rodríguez-Pose, 2018)—is stirring up resentment among their inhabitants, who are turning against the system at the ballot box (Rodden, 2019; Wuthnow, 2019).

There is some overlap between interpersonal inequalities and economic decline at the subnational level. The destruction of low- and middle-skilled jobs, as a consequence of outsourcing and offshoring, mainly affects less formally educated workers, many of which live in already declining manufacturing areas (Autor et al., 2016). Meanwhile, the benefits of the process are reaped by skilled workers living in dynamic places (Gagliardi et al., 2021). Nevertheless, these remain distinct phenomena. Many places where economic growth has been stagnant in the last two decades are less unequal than superstar cities, where there is a greater concentration of both high-income individuals, on the one hand, and precarious low-skilled service workers, on the other.

The rise in inequality is at the centre of most economic explanations of the increase in populist voting. Therefore, it is surprising that comparative analyses of the extent to which differences in and the growth of both interpersonal and interterritorial inequality drive the populist momentum remain rare. The few studies that have considered the link between interpersonal inequality and populist voting [e.g. Becker et al. (2017) for the case of Europe and Mutz (2018) for the USA) treat the issue of interpersonal inequality almost as a sideshow to other relevant explanations behind the rise of populism. These contributions find limited evidence of a link between individual economic hardship and the rise of populist vote. Research connecting the economic trajectories of regions and right-wing populist voting suggests a strong link between long-term economic and demographic decline and the choice of populist options at the ballot box. This is reflected in purely descriptive studies (Broz et al., 2021) and in those relying on econometric models. The factors behind this link vary between Europe—GDP and industrial decline (Dijkstra et al., 2020)—and the USA—employment and population decline (Rodríguez-Pose et al., 2021). However, these studies have not dwelt on the rise of interpersonal inequalities at the local level.

Hence, considerable gaps remain regarding the potential link between levels and the growth of local interpersonal inequality and the reaction of those in left-behind regions at the ballot box. In particular, there is, so far, no analysis at a European level linking interpersonal inequality across regions with the rise of populism. This is because interpersonal inequality data were not readily available at a subnational level for most European countries. Hence, the question of whether it is left-behind or more unequal places (or both) driving the recent surge in right-wing populism remains open.

This article addresses this gap by systematically looking at differences in interpersonal inequality and in regional wealth at a local and regional level at the beginning of the twenty-first century. It does so by using existing interpersonal inequality data at county level in the USA and putting together the first comprehensive dataset of regional (NUTS3) interpersonal inequality for the member states of the European Union (EU), plus the UK. These datasets allow us to examine changes in inequality—using the Gini coefficient—and local economic growth as potential elements connected to votes for populist parties in national elections. To the best of our knowledge, it is the first study to assess the relationship between left-behind places, interpersonal inequalities and populism systematically and at a fine geographical scale for Europe and the USA.

The article also analyses whether the impact of interpersonal inequality and regional economic growth is intertwined with immigration and the racial composition of places. This is because these main cultural factors have been highlighted as central drivers of the rise of populist vote. Especially for the far right, voter anxiety goes beyond material needs and reflects the feeling of losing a dominant status in society, often at the expense of other social groups, notably immigrants, minorities and urban educated elites (Mutz, 2018). Cultural and economic factors can reinforce each other, leading to ‘losers’ in multiple dimensions. However, empirical research estimating the joint effect of economic and racial/immigration dimensions remains limited.

This article moves forward this debate by interacting our measures of territorial decline and local interpersonal inequalities with the racial composition of counties in the USA, where race has played a vital role in the support for Donald Trump (Mutz, 2018; Hinojosa Ojeda and Telles, 2021). For Europe, where data on racial composition are not available, we interact these economic factors with the share of individuals born in the country.

The results show that inequalities and economic decline, coupled with racist and anti-immigration attitudes, are powerful predictors of populism, although there are differences between the USA and Europe. In line with previous studies (Dijkstra et al., 2020), populism has taken root in European regions undergoing long-term economic decline. This is particularly the case of areas where immigration is high, suggesting that anti-immigration discourses take hold mostly in places where economic conditions are deteriorating. In the USA, racial composition moderates the effects of economic factors. Economic decline and the growth of interpersonal inequalities are linked to Trump support, but only in places with a high share of white residents. In contrast, the same factors have pushed the electorate of racially diverse areas towards the Democrats. In short, racial cleavages in the USA become even more salient where income polarisation is on the rise and economic growth on a relative declining path.

The article is structured as follows. The first section discusses the impact that the ‘new’ post-industrial economy has had on the evolution of economic output and inequalities at the local level. This provides the lens through which to understand how local economic conditions may trigger populist support. The second section delves into scholarly analysis linking inequalities, economic decline and how both factors, in turn, are connected to geographical patterns of voting behaviour. This is followed by the introduction of the model, the data and the methods. The results of the empirical analysis are presented and discussed in a subsequent section, before concluding and discussing some caveats of the analysis.

2. Economic change and inequality

2.1. The local impact of economic change

The rise in both interpersonal inequalities and the stagnation of certain territories in recent decades are intrinsically connected with a major transformation in the global production system. Production in developed economies has shifted from an ‘old’ industrial system with geographically dispersed activities to a ‘new’ tech economy, increasingly concentrated in big urban agglomerations. Leading ‘old’ economy firms relied on hiring large numbers of workers in plants that were scattered across regions of the developed world. These Fordist firms guaranteed that not only the rise of interpersonal inequalities was kept at bay, but also that territorial inequality kept on declining (Carrascal-Incera et al., 2020; Rosés and Wolf, 2021). In contrast, ‘new’ economy firms operate under very different conditions. Giant tech firms have virtually no supply chains. With just a fraction of the workforce, they can interact directly with customers on a global scale (Teece and Linden, 2017, 2). Hence, from a territorial perspective, the ‘new’ model leads to the concentration of highly innovative and profit-generating economic activity in the hands of highly skilled workers, who flock to large metropolises. This undercuts the supply and customer chains that were the bread-and-butter of poorer and often less dense regions both in Europe and the USA.

The consequences of this massive transformation in the economic model are becoming plain to see. First, there is a substantial impact on interpersonal inequality. Technological progress and the shift in economic and industrial organisation have changed who and how people benefit from the system. Fewer workers with a more homogeneous educational background are reaping the lion’s share of benefits, in contrast to a previously more widespread and equal distribution of economic activity. More constrained value chains limit multiplier effects on the broader economy. The result is a rise in interpersonal inequality, characterised in the developed world by the stagnation and relative decline in income of the middle- and lower-middle classes (Milanovic, 2016; Dorling, 2019). This process is driving social and political tensions (Milanovic, 2010; Piketty, 2013).

Interpersonal inequalities have been extensively studied at the national level. Yet, when it comes to voting outcomes, the phenomenon is likely to be more relevant at the local level. Individuals evaluate their own position relative to those around them, drawing on their daily experiences and interactions, which take place in their local context (Cruces et al., 2013; García-Castro et al., 2020). As a result, places where interpersonal inequalities are high and/or on the rise—and where, consequently, the contrast between ‘winners’ and ‘losers’ has become more salient—have frequently been targeted by populist leaders.

Simultaneously, the concentration of economic output in large agglomerations has led to greater inequality between places (Carrascal-Incera et al., 2020; Rosés and Wolf, 2021; Feldman et al., 2021). Territories that previously attracted manufacturing or provided many of the small and medium-sized firms that formally supplied ‘old’ industry conglomerates or delivered goods and services to their workers in medium-size cities, towns and rural areas have entered an irretrievable decline. The concentration of high-end economic activity in large urban agglomerations and superstar cities (Kemeny and Storper, 2020) and the parallel demise of many low- and medium-tech plants outside these large agglomerations has left many areas wedged in a development trap (Iammarino et al., 2019; Diemer et al., 2022). They are stuck between a rock and a hard place: incapable of competing in high-end production with the large agglomerations within their countries or other parts of the world, but also in low-tech production with emerging markets (Vandermotten et al., 1990). The decline of economic activity in these areas is leaving large swaths of rural regions, but also towns and medium-sized cities across the developed world behind, giving the growing impression that these places ‘no longer matter’ in the more integrated and globalised ‘new’ economy (Rodríguez-Pose, 2018).

2.2. Explaining populism through inequality

Economic explanations for the rise of populism have been centred around the rise in inequalities brought about by technological progress, the shift in the production system and global economic integration. This has been a tale of two inequalities.

On the one hand, the economic explanations on the rise of right-wing populism have mostly focused on the increase of interpersonal inequality. The greater resentment at a system that benefits the elite, while failing large layers of the population and rendering them vulnerable, is at the heart of increasing dissatisfaction with traditional political options.

Economic changes linked to trade and economic integration have rendered many types of jobs redundant, destroying middle-skilled jobs and provoking fears about workers’ well-being (Frey and Stutzer, 2002). Greater trade with the rest of the world has also resulted in what is known as the China shock (Autor et al., 2016; Colantone and Stanig, 2018, 2019); that is, an incapacity to compete with the production of goods and services delivered at lower costs by China and other emerging economies. The rising competitiveness of the Global South and the offshoring of more routine and, increasingly, technologically advanced activities from developed countries have left many individuals, who depended on the types of jobs under threat, economically vulnerable (Guiso et al., 2017; Gagliardi et al., 2021). At the same time, the concentration of high-income, high-skilled workers in the services sectors, particularly in tech clusters, is fostering increasing employment polarisation in urban agglomerations (Kwon and Sorenson, 2021).

The resulting job insecurity (Algan et al., 2017) and interpersonal inequalities (Piketty, 2013; Milanovic, 2016) arising from these economic processes have been skillfully exploited by populist politicians (Rodrik, 2018). Decreasing social mobility coupled with accumulation of wealth in a dwindling number of individuals is resented by voters as a proof that the system is working only for the elites, who are mostly concerned about protecting their position. Redistribution and economic egalitarianism have featured prominently in the programmes of left-wing populist parties, such as Podemos (Marcos-Marne et al., 2020). But far-right candidates have also embraced discourses against the wealthy liberal elites. The financial elites, international bureaucrats and big businesses have been portrayed by populist leaders, right and left, as the promoters of an unfair economic system (March, 2017).

Finally, what many feel has been an inappropriate response to the 2007–08 Great Recession has also made more salient the failure of the current system in redistributing income and protecting the poor. The implementation of austerity measures, fundamentally in Europe, has been thought to stifle the recovery, striking particularly hard those more affected by the financial crisis (Gray and Barford, 2018; Fetzer, 2019). A loss of opportunities has ensued, often combined with the expansion of precarious work as the last resource for vulnerable individuals (Gidron and Hall, 2017; Pástor and Veronesi, 2021). The perception that the measures implemented after the crisis led to an unequal recovery, widening the gap between the haves and the have-nots (O’Connor, 2017; Eichengreen, 2018; Rodrik, 2018; Pástor and Veronesi, 2021) is another driver of discontent in the developed world. Those whose opportunities and chances of making social and economic progress have been taken away have become prone to voting against the system, regardless of where they live.

On the other hand, explanations of the rise of discontent have sought to link it to the prolonged economic and/or demographic decline of certain places and the related rise in interterritorial inequality (Essletzbichler et al., 2018; Martin et al., 2018; Rodríguez-Pose, 2018). Globalisation has deepened territorial disparities across the world and within countries. The decline or demise of many activities that once were dominant has meant that many of the traditional centres of the industrial revolution have lost their former edge. They have been plunged into a prolonged economic and often demographic decline. The economies of the Rustbelt in the USA, the traditional industrial hubs of Wallonia in Belgium, the North of England, north-eastern France or many industrial districts of the North of Italy, face considerable difficulties in transitioning to a knowledge-based economy. Places facing higher import competition from foreign markets are more likely to vote against incumbent governments (Jensen et al., 2017) and for populist parties (Colantone and Stanig, 2018, 2019). Similarly, former industrial, but also many rural areas undergoing population decline have been losing out in terms of the services on offer for their citizens (Guilluy, 2018; Collantes and Pinilla, 2019). A highly regionalised recovery from the Great Recession is also contributing to discontent, with regions highly exposed to austerity more inclined to support populist alternatives at both extremes of the political spectrum (Artelaris and Tsirbas, 2018).

The inhabitants of these regions have lost out. Their opportunities are limited, and they often feel stuck in places that ‘don’t matter’ for decision-makers (Rodríguez-Pose, 2018). Places from where they cannot or are unwilling to move, even though their prospects for personal and collective development have flatlined (Lee et al., 2018). They also frequently feel unable to influence policy decisions. This is the relentless rise of a ‘geography of discontent’ (Dijkstra et al., 2020; McCann, 2020), which has political consequences. Regions suffering from different forms of economic decline have embraced far-right populist parties. This is the case for areas in which population and employment have declined in recent decades [Rodríguez-Pose et al. (2021) for the USA and Broz et al. (2021) for the USA and Europe]; regions with strong GDP and industrial decline [Dijkstra et al. (2020) for Europe]; areas where real-estate property has lost value [Adler and Ansell (2020) for the UK] and places where manufacturing employment has decreased [Baccini and Weymouth (2021) for the USA].

These regions have become the epicentre of deepening division and of a new, angrier form of politics. They have said ‘enough is enough’ and are intent on bringing down a system that they perceive as leaving behind their communities. This sentiment underlines a double failure (Frieden, 2019): the government has failed to compensate them for the economic losses derived from an ‘unfair’ globalisation, and mainstream parties have failed to bring their concerns onto the political agenda. The result is a shifting towards radical, far-right populist voting, often at the expense of their own prosperity (Los et al., 2017; Guilluy, 2018; Rodríguez-Pose, 2020).

The rise of interpersonal inequalities and the emergence of ‘pockets’ of left-behind regions are both phenomena stemming from recent changes in the dominant production model in advanced economies and, hence, cannot be considered as completely independent. Indeed, the decline of manufacturing may foster inequality through the destruction of employment, while individuals suffering from greater income shocks and precarious employment can become concentrated in specific geographies. However, increasingly unequal regions and those with stagnant growth are not necessarily the same.  Annex 1 shows the correlations between increasing inequalities and GDP growth across places in Europe and the USA. The correlation is limited for the EU and non-existing for the USA.

2.3. Interpersonal and interterritorial inequalities as drivers of populism

Although the rise in political discontent and resentment has attracted considerable scholarly attention, there is little research considering the distinction between interpersonal or interterritorial inequality and the surge in far-right populism, or populism more broadly. Most research has concentrated on the study of the individual characteristics of disgruntled voters. Factors such as gender, age, race, level of education and personal wealth—with data mainly from surveys—have featured prominently (e.g., Goodwin and Heath, 2016a; Hobolt, 2016). Some geographical aspects, such as density and the rural–urban divide, have also been explored (Cramer, 2016; Rodden, 2019; Wuthnow, 2019; de Dominicis et al., 2022). Others, such as the long-term economic, social and demographic trajectories of territories, have started to be explored only recently (Dijkstra et al., 2020). Moreover, the link between the rise in inequalities provoked by the shift in the economic and production models of recent decades with the rise of discontent and resentment still needs to be developed.

There is, so far, limited systematic subnational analysis about to what extent and how the changes linked to the ‘new’ economy cause discontent (e.g., Colantone and Stanig, 2018; Los et al., 2017; Artelaris and Tsirbas, 2018; Rodríguez-Pose, 2018; Adler and Ansell, 2020; Dijkstra et al., 2020; Dorn et al., 2020; Patana, 2020; Broz et al., 2021; Baccini and Weymouth, 2021). The country remains frequently the main, rather coarse, territorial unit of analysis. However, support for anti-system parties is uneven within countries—often more than across countries—suggesting that the action happens at a much finer geographical scale: at the level of regions or even communities, localities or electoral districts. In fact, in contrast to analyses at a finer geographical scale, country-level studies often struggle to find a correlation between economic phenomena, such as openness to trade and globalisation and support for populist parties (Bergh and Kärnä, 2020). This highlights the need to do comparative analyses that reduce the degree of aggregation and delve into how economic changes affect the life and the well-being of individuals wherever they live.

Moreover, most research on populism in the developed world has been concerned with individual countries. However, a considerable gap remains in terms of comparative analyses involving many countries. Research considering populism as a European-wide phenomenon is slowly gaining ground (e.g. Bergh and Kärnä, 2020; Dijkstra et al., 2020; Nowakowski, 2021). However, comparative, quantitative analyses of the drivers of populism at a subnational level in Europe and the USA are still uncommon. Although Broz et al. (2021) study some economic factors behind the rise of populist voting in both continents, their analysis relies exclusively on stylised facts. Hence, more detailed econometric exercises are needed to determine what is really behind the discontent at the ballot box on both sides of the Atlantic.

More importantly for the purpose of this article is the fact that virtually no research has confronted interpersonal and interterritorial inequality. The strands of literature dealing with interpersonal inequality, on the one hand, and interterritorial inequality, on the other, have run in parallel with limited cross-fertilisation. Research on interpersonal inequality has also often been constrained to studying inequality at the country level (e.g. Pástor and Veronesi, 2021). One of the potential reasons has been the lack of suitable interpersonal inequality data, especially in the case of Europe. This is unfortunate since the social psychology literature suggests that individuals form their perceptions of economic conditions, including the degree of inequalities, through their daily experiences, which take place at the local level. For instance, individuals assess their income level based on a sub-sample of the population and generalise over the whole society, leading to biases when assessing their relative wealth at the national level (Cruces et al., 2013). Moreover, people experiencing greater inequalities in their immediate social (García-Castro et al., 2020) and geographical (Larsson et al., 2021) context find inequalities less tolerable. This further highlights the need to incorporate the subnational level as a relevant geographical scale for the study of economic conditions.

When research has considered inequality as a driver of populism, the results are inconclusive and often do not match either the theory or the dominant narrative about the rise of populist voting. Becker et al. (2017), for example, found that their measure of interpersonal inequality—the interquartile pay range—was either irrelevant for or had, in contrast to expectations, a negative relationship with the Brexit vote. Others, however, found a connection between relative income decline and the support for radical parties (Burgoon et al., 2019). As already discussed, the evolution of territorial inequalities, in particular different measures of economic and demographic decline—GDP and industrial decline in the case of Europe and employment and demographic decline for the USA—are connected to populism (Dijkstra et al., 2020; Rodríguez-Pose et al., 2021). Overall, however, we remain in the dark as to whether it is mainly interpersonal inequality or regional decline that drives the shift from mainstream to populist parties at the ballot box.

2.4. Race, immigration and the link with economic explanations of populism

The role of race and immigration is at the centre of cultural explanations for the rise of far-right populism. The rise of more diverse, multicultural and cosmopolitan societies has estranged layers of the population from a world they feel ill-at-ease with (Norris and Inglehart, 2019), almost making them ‘strangers in their own land’ (Hochschild, 2018). In this regard, the outsider—often the immigrant or racially different—is seen as a threat to the ‘right’ way of life and to the supremacy of the dominant group. This element has also a clear regional component. For instance, Luca et al. (2023) show that anti-immigration attitudes are more prevalent in rural areas.

In the case of the USA, Mutz (2018) has argued that support for Donald Trump reflects the anxiety of white, high-status groups, fearing losing out to cultural minorities rather than objective decline. White individuals are more likely to vote for Trump and racist resentment is a key factor behind the support for the Republican leader (Hooghe and Dassonneville, 2018). The fear of difference is so strong it prevails over objective realities. For instance, Hinojosa Ojeda and Telles (2021) found that anti-immigration attitudes are linked to support for Trump, while actual immigration is not.

Immigration has constantly been used by populist parties on the far-right to rally the support of voters. Far-right populists have exploited the narrative of immigration as a threat to the Western way of life and the common achievements of liberal democracies (Yilmaz, 2012). In Europe, in particular, Muslim immigrants have been portrayed as a national security risk in the context of terrorist attacks (Hogan and Haltinner, 2015), while the rise of immigration has been described as a ‘controlled invasion’ (Naxera and Krčál, 2018). On the economic front, foreign immigrants have been accused of depressing wages, abusing social welfare and stealing native workers’ jobs.

Despite the anti-immigration rhetoric of far-right populist parties, the empirical evidence on immigration and support for populist parties is mixed. Some authors find that the levels of immigration increase populist voting (Otto and Steinhardt, 2014), others find significant only its rise (Kaufmann, 2018; Patana, 2020) or even that high levels of immigration can reduce populist voting (Della Posta, 2013; Charitopoulou and García-Manglano, 2018).

The literature on economic and cultural factors has run in parallel, often in disagreement. This is even though individuals can be ‘losers’ of the system on several dimensions, and that cultural and economic anxieties are likely to reinforce each other. In most cases, empirical studies have pointed out that both economic factors and racial and immigration attitudes matter for populism. For instance, Rodrik (2021) showed that those switching from the Democrats to Donald Trump held more negative views on immigration and the rights of minorities, but also suffered from higher financial distress. Similarly, Clarke et al. (2017) found that both voters who perceived that their economic situation would improve after leaving the EU and those who wanted a decrease in immigration were more likely to support Brexit. It has also been found that individuals who endure hard economic conditions and have strong anti-immigration sentiments are more likely to vote for radical right parties (Rooduijn et al., 2021).

Fewer studies have addressed how economic factors mediate anti-immigration sentiment and vice versa. As the financial security of families decrease, members are more likely to resent immigrants, hence activating support for populist alternatives (Ferrari, 2021). However, there is no evidence on how economic factors, such as inequality and local economic decline, interact with the demographic composition of territories. By evaluating the joint effects of our economic variables and race (USA) and immigration (EU), we address an important gap in existing knowledge.

3. Model, data and method

3.1. Model

To assess if the surge in populism in Europe and the USA is a factor of the levels and growth in interpersonal and/or interterritorial inequalities—or both—we conduct the analysis at a fine territorial level: small regions (NUTS3) for Europe and counties for the USA. This choice of territorial units allows us to zoom into whether it is left-behind or more unequal places that shape the electoral map of populism in both continents.¹

We adopt the following econometric model:

where $Populist vote$ represents the vote for far-right populist parties in region i of country c in the last national election during the period of analysis t. This vote is measured as the Trump margin in the US case and the share of vote for far-right populist parties in Europe. Interpersonal denotes the level of interpersonal inequality in territory i at time t; $ΔInterpersonal$ represents the growth in interpersonal inequality in a particular place i between the turn-of-the-century (k) and the end of the period of analysis (t); $RegionalWealth$ is captured by GDP per capita (EU) and income per capita (USA) in region i at time t; $ΔRegionalWealth$ is the growth in wealth in that particular region between k and t; $X¯$ represents a vector of variables identified in previous research as factors driving populism; $ν$ captures country fixed effects for Europe; $ε$ is the error term; $α$ is the intercept and $β$ and $γ$ are the coefficients.

The analysis is expanded to include the interaction between the economic factors and race and immigration. In the USA, we choose the share of white, non-Hispanic residents, as technological change and racism in the USA are intertwined not only to trigger economic decline, but also to propel the rise of Trumpism (Hackworth, 2019). Moreover, scholarly literature for the USA has highlighted racist sentiment as a prominent factor in Trump support (Hooghe and Dassonneville, 2018). In Europe, where consistent and comparable data on racial composition at the regional level are not available, we resort to the share of residents born in the country, that is, non-immigrant population, given the literature underscoring the role of immigration in populist support (Otto and Steinhardt, 2014; Kaufmann, 2018).

3.2. Data

3.2.1. Far-right populist parties

The dependent variable, $Populist votei,c,t$⁠, captures the share of votes in national elections for parties considered as far-right populists. Populism is a highly contested concept, under increasing scrutiny in the scholarly literature. Mudde (2004, 543) defines populism as a thin-centred ideology that considers society divided into two homogeneous and antagonistic groups: ‘the pure people’ and ‘the corrupt elite’. Populist parties adopt a discourse that blames the ‘elite’ for illegitimately creating and sustaining a system that exploits and leaves the ‘people’ behind. The solution offered involves overriding the checks and balances of representative democracy and the establishment of a system in which the general will of the ‘people’, incarnated in the populist leader, becomes the legitimate mechanism for policymaking. Far-right populists embrace this basic framework and fill it with radical right principles, including nationalism, anti-immigration and social conservatism (Akkerman, 2003).

For the USA, far-right populist voting is captured by the Trump margin, that is, the swing in the share of votes towards the Republican Party between the 2012 presidential election—when Mitt Romney was the Republican presidential candidate—and the 2020 election, at county level. Although Donald Trump is a businessman from the American ‘establishment’, he has deployed a populist rhetoric that clearly distinguishes him from traditional mainstream Republican candidates. His presidency was characterised by the rejection of both national and international institutions as legitimate sources of sovereignty (Broz et al., 2021), a rejection that was taken to the extreme by his contestation of the 2020 presidential election results and the US Capitol assault by his followers. This authoritarian narrative is interwoven with an anti-establishment discourse, in which economic, political and bureaucratic elites are made responsible for the supposed decline of the country. He also oversaw a ‘cultural backlash’, prompted by his strong nationalist and anti-immigration positions (Norris and Inglehart, 2019), of which the Mexican wall is the ultimate example.

For Europe, we consider populist parties those at the extreme right of the political spectrum in the Chapel Hill Expert Survey (CHES) 2014–2017. The CHES classifies the overall ideology of a party on a scale from 0 (left) to 10 (right). We select those parties scoring 8.5 or above, which can be considered radical far-right parties. These parties also score very high in another CHES scale in terms of anti-elite rhetoric. This cut-off is a methodological choice, but it roughly corresponds with those parties considered far-right populist by media outlets and the academic literature (e.g. Rassemblement National in France, the Lega in Italy, Jobbik in Hungary, the Freedom Party in Austria and the AFD in Germany). The share of valid anti-elite votes is gathered in national legislative elections taking place between 2014 and 2018. These data were extracted from the Constituency-level Elections Archive dataset (Kollman et al., 2016). It must be borne in mind that, whereas the US data cover the share of votes for the Republican Party in a comprehensive way, not all parties and, for that sake, not all far-right populist parties in Europe are included in the CHES. The phenomenon of far-right populism has only taken off recently in many European countries. Whereas far-right populist vote in national parliamentary elections in the EU represented a mere 4% of the total in 2009, its share rose to 26% in similar elections in 2022. In several EU countries, the support for far-right parties was negligible until a few years ago and in some it remains almost inexistent. This means that—considering that we also measure populism in the USA using the Trump margin relative to the 2012 election—it is more fitting to analyse levels than change in support for right-wing populism.

3.2.2. Independent variables and controls

The independent variables of interest measure wealth—proxied by income per capita for counties in the USA and GDP per capita for regions in Europe—and interpersonal inequality, as well as their respective evolution over time.

For the USA, the income per capita (in 1,000$) data are collected for 2000 until 2019 from the Bureau of Economic Analysis (BEA). 2000 is used as the baseline. We compute the average annual real growth in income per capita over the period. Interpersonal inequality is measured by the Gini coefficient extracted from the US Census Bureau. It measures the within-county inequality at household level in 2000 and 2019. We also calculate the growth rate between both years.

For Europe, regional wealth is measured using the GDP per capita in purchasing power standards (in 1,000 s EUR). The source is the Annual Regional Database of the European Commission’s Directorate General for Regional and Urban Policy (ARDECO) for 2000 and 2014. We calculate the average annual real GDP per capita growth between both years. As for interpersonal inequality, we put together the first dataset on interpersonal regional inequality for NUTS3 regions. We rely on two sources—depending on availability for each country—offering comparable inequality data: the Luxembourg Income Study (LIS)² and the European Union Statistics on Income and Living Conditions (EU-SILC). We compute the Gini coefficient for equivalised household disposable income in 2004 and 2016³ and calculate the growth between both years. Inequality data are not available for the reference years in all countries. When this is the case, we use the data for the nearest available year.⁴

The analysis also includes several control variables under Vector $X¯i,t$⁠. These represent factors that, according to the extant literature, influence the rise of far-right populist voting on both sides of the Atlantic. The control variables are:

• The share of population over 65 (Europe) and 55 (USA), given that the elderly population has been more prone to support far-right populist parties (Eatwell and Goodwin, 2018; Rodrik, 2018).

• The level of education, proxied by the share of adults (USA) and the share of working-age population (Europe) with tertiary education (Goodwin and Heath, 2016a; Jump and Michell, 2020).

• The degree of rurality, given the stark urban–rural contrast in social values across the developed world (Luca et al., 2023) and the observed urban–rural political polarisation in the USA (Rodden, 2019). This is captured both by the population density (ln) and the share of rural inhabitants in Europe, and the most up-to-date rural–urban continuum code of the US Department of Agriculture, which classifies US counties in values ranging from 1 (metropolitan areas) to 9 (completely rural counties).

• The employment rate of the working-age population in European regions and the unemployment rate in US counties, given the literature linking the absence of (Chen, 2020; Passari, 2020) or the decline (Rodríguez-Pose et al., 2021) in jobs to the rise of populism. Employment rates for Europe at the NUTS 3 level are approximated using total employment data from regional accounts, divided by population data for the age range 17–74 at the regional and metropolitan region levels in 2015.

• The share of the population born in the country, as the presence of migrants and a strong anti-immigrant discourse has been associated with far-right support at the ballot box, both in Europe (e.g. Otto and Steinhardt, 2014) and the USA (Hinojosa Ojeda and Telles, 2021).

• The share of the white, non-Hispanic population at county level in the US regressions, since race has been identified as a factor behind US populism (Hooghe and Dassonneville, 2018; Kaufmann, 2018; Mutz, 2018). Equivalent data are not available for European regions.

• Finally, for Europe, the share of votes for parties not classified in the 2017 CHES. A fraction of votes, representing on average 5% of all votes in national elections, are not monitored by the CHES. These are mainly votes for independent parties running on local issues and tickets or parties that were created after 2017.

Of the control variables defined above, the share of white non-Hispanic population (USA) and the share of share of population born in the country (Europe) are interacted with our economic factors.

A list of all the variables included in the analysis and their sources for the USA and Europe are presented in Tables A1 and A2 in  Annex 2.

3.2.3. Method and geographical coverage

The analysis is performed by means of variations of ordinary least squares estimations of model (1), with robust standard errors clustered at regional level in the case of Europe, and at state level for the USA.

The geographical coverage involves NUTS3 regions in Europe⁵ and counties in the USA. For the USA, we follow the 2017 county division of the BEA. Alaska is excluded and adjustments were made to the county composition in the state of Virginia, to account for boundary changes. This leaves a total of 3067 counties. For Europe, NUTS3 regions are the smallest territorial unit in the EU’s classification, encompassing the regional tier right above the local and/or municipal level. We exclude from the analysis those countries where there were no parties classified as populist using CHES criteria running in national elections between 2014 and 2018. As inequality data at the regional scale are not available for all European countries, the analysis is limited to 20 EU member-states,⁶ plus the UK. Due to some transformations in the regional units over time, we cannot recover data for certain NUTS3 regions. This leaves us with a maximum of 812 regions.

NUTS3 regions and US counties are different local units. Sizes vary considerably both between categories and within them, which limits their comparability. Nonetheless, both units represent the lowest level at which data are available on both sides of the Atlantic, and hence the one that more closely can match the local reality of individuals.

4. Results

4.1. Descriptive analysis

Before delving into the multivariate analysis, we perform a descriptive analysis of bivariate correlations between our main variables of interest and far-right populist voting in the USA and Europe.

Figure 1 shows the correlations for the USA. We plot the levels of and the growth in both local income per capita and interpersonal inequality against the Trump margin in the 2020 presidential election. We also include the correlation between the Trump margin and our race variable of interest, that is, the share of white, non-Hispanic residents.

As Figure 1 suggests, there is no clear correlation between the growth in inequalities and county wealth and the Trump margin. Previous research already signalled that it was population and employment decline, rather than income, which correlated with the rise of Trump (Rodríguez-Pose et al., 2021). Places where interpersonal inequality is higher and particularly richer places drifted apart from Donald Trump in the 2020 elections. This is not surprising, since more unequal and prosperous counties are likely to be urban. Urban counties behaved mostly as liberal strongholds in the 2020 presidential elections. As expected, there is a positive relationship between the share of the white, non-Hispanic population and the Trump margin, although this seems to be driven by areas with a very high proportion of whites. There are also some strong outliers in places with a very high racial mix.

Figure 2 shows the correlations between our main economic variables, the share of non-immigrant population, and the vote for far-right populist parties in Europe. This descriptive exercise should be treated with caution. There are considerable variations in the share of far-right populist voting across countries, which are not accounted for in these correlations. Two strong correlations emerge from this simple exercise. In line with previous research, populism has been successful in areas where economic growth has been stagnant (Dijkstra et al., 2020). This is also the case in regions where interpersonal inequalities have increased the most over the period. The remaining relationships are quite muted, with almost flat fitted lines.

4.2. Multivariate analysis

To address our main research question, we use an OLS multivariate analysis of model (1). Four different regressions are presented in each case. As we are interested both in the levels and growth of our main economic variables, regression (1) includes the levels of regional wealth and interpersonal inequality and regression (2) focuses on the change in these variables. To account for potential confounders, regression (3) includes the four economic variables together with a set of controls. Lastly, the final regression includes the interaction between our main economic variables and race (USA) and the share of non-immigrant population (EU) (4).

The regressions for the USA are presented in Table 1. Prior to including the controls, we observe that richer and more unequal counties swung less to Donald Trump, while the growth of inequalities and in income per capita are not related with the Trump margin (Table 1, Regressions 1 and 2). This relationship is not trivial: an increase of 1,000$in the income per capita of US counties is associated with an approximate decrease in the Trump margin of 0.2 (the median Trump margin in 2020 was 5.5). Similarly, a one point increase in the Gini coefficient is related to a decrease of 0.25 in the Trump margin. The Trump vote is concentrated in rural areas (Rodden, 2019), which are in general more equal but poorer than urban agglomerations. Wealthier areas also have more educated, younger and more racially diverse populations, all factors associated with a Democratic leaning. This hypothesis is confirmed by the inclusion of controls, which render all economic variables insignificant (Table 1 and Regression 3). The support for Donald Trump in the 2020 election was indeed larger in rural counties, as well as in places with a higher share of highly educated, older and white residents.

These results go in line with cultural explanations of populism, particularly those highlighting the status-loss anxiety of white local majorities (e.g. Mutz, 2018), particularly manual workers, and racial resentment (e.g. Hooghe and Dassonneville, 2018), as the main determinants of populist vote. However, this explanation misses half of the story. The results in Regression (3) mask strong differences in the importance of economic factors for counties with different racial compositions. Regression 4 portrays the joint impact of race and the economic variables in Donald Trump's margin. The interaction is significant for three of these economic factors: interpersonal inequalities; inequality growth and income per capita growth.

To unpack the meaning of these interactions, Figure 3 plots the predictive margins of the model (Regression 4) at various levels of the economic variables, for counties with different degrees of ‘whiteness’ (percentiles in parenthesis).⁷ The figure suggests that interpersonal inequality and economic decline polarise the vote between white and non-white areas. Voters in white-majority areas where inequalities grew the most during the period of analysis swung towards Trump. Those in areas where inequalities rose, but which are racially diverse, leaned more towards Joe Biden. Voters in counties where interpersonal inequalities declined or stagnated voted more alike, regardless of the racial composition of the county. A similar pattern is observed when considering annual growth in income per capita. The Trump margin in low-growth, white-majority counties was significantly larger than in the average county. Meanwhile, more racially diverse, economically declining counties barely experienced a swing towards Trump in the 2020 presidential elections.

These results suggest that in the USA, economic and racial anxieties are intertwined. A worsening economy and rises in inequality boosted Trump support in white-majority communities. In those areas, Trump’s anti-immigration discourse, coupled with a rhetoric of making America great Again by bringing back jobs to white impoverished workers, was bought by voters as the solution to economic stagnation and rising inequalities. In contrast, racially diverse counties where economic conditions have deteriorated were prone to see Trump’s anti-minority stand as a further threat to their already vulnerable position. These counties reacted against Trump’s populist discourse more strongly than the generally more prosperous, comparable white-majority communities.

A different pattern emerges when analysing the levels of interpersonal inequality. The rise in discontent due to high interpersonal inequality in US cities may be reflected in the rise and strength of social movements, such as Occupy Wall Street, but not in more support for Trump’s populism. The gulf between the richest and the poorest US citizens, often happening within the urban microcosm of big cities, is large. However, the inhabitants of the upmarket Manhattan districts and those of the impoverished areas of the Bronx have continued to vote together (Rodríguez-Pose, 2020). Rather, the swing towards Trump was stronger in less unequal, but largely white-majority areas.

Table 2 presents the results for Europe. As in the case of the USA, the correlation between the level of inequalities and the support for populism is negative (Table 2 and Regression 1). This result still holds after including a battery of economic and demographic controls (Regression 3). More unequal places have not rejected mainstream candidates. In the USA, white areas where income polarisation increased the most became voting grounds for Donald Trump. In Europe, there also seems to be a correlation between places where inequality has risen and the support for far-right parties (Regression 3). Populist leaders have been able to capitalise on the role of inequalities, but only in those areas where these have been on the rise.

The main economic explanation behind the populist vote in Europe is the economic performance of the region, measured as annual regional growth in GDP per capita (Regressions 3 and 4). The effect is significant in all regressions and quite substantial: a one percentage point increase in annual GDP per capita growth over the period of analysis (2000–2014) is linked to a decrease of almost half a percentage point in the share of populist voting (Regression 3). Our results go in line with previous research documenting the relevance of left-behind regions as strongholds of support for populist alternatives across Europe (Dijkstra et al., 2020). The experience of collective loss, coupled with the perception that national administrations have abandoned the needs of left-behind regions, seems to be powerful drivers of populist sentiment (Rodríguez-Pose, 2018).

More broadly, the results suggest that it is the worsening of economic conditions over time—rather than low levels of GDP per capita—what brings voters to embrace far-right populist alternatives. In this regard, our findings resonate with the work of those that perceive populism as a reaction to a loss in economic prospects [Adler and Ansell (2020) for the UK; Baccini and Weymouth (2021) for the USA and Burgoon et al. (2019) for Europe].

We do not have data on racial demographics for European regions, so in this case we interact our main economic variables with the share of those born in the country (Regression 4). The only interaction that is significant is that between economic decline and the share of non-immigrant population. To facilitate the interpretation, Figure 4 plots the predictive margins of the model (Regression 4) at various levels of GDP per capita growth, for regions with a different share of non-immigrant population (percentiles in parenthesis).

A comparison between the USA and Europe is not possible, as most immigrants do not vote in Europe. However, at least one interesting similarity emerges: identity politics seem to matter only in places where the economy has deteriorated. Indeed, according to our model, the vote for populist alternatives is concentrated in areas with higher levels of immigration, but where annual GDP per capita growth has stagnated. The most plausible interpretation is that immigration triggers support for far-right populist parties only in areas where local population see the presence of immigrants as a threat to their economic prospects. In fact, regions with high levels of immigration but also higher economic growth have strongly rejected populism.

Most controls in regressions 3 and 4 go in line with previous studies. For instance, lower levels of education correlate with a higher share of votes for far-right populists. Similarly, these political alternatives are also stronger in areas with an older population. Interestingly, populist leaders tend to gather more support in regions where employment rates are higher. However, the data do not allow us to identify the quality of employment and this seems to be particularly relevant for the emergence of such parties (Im, 2021; Zagórski, 2021).

5. Conclusion

Inequality has been a cornerstone of explanations for the recent explosion of populism in developed countries. Economic explanations of the Brexit vote, the rise of populism in different countries of Europe or the election of Donald Trump as president of the USA frequently put greater economic vulnerability and the progressive demise of the middle classes at the centre of the populist phenomenon. Territorial inequality explanations have also made considerable inroads in recent years. However, in most of these analyses, a divide existed between theory and narrative, on the one hand, and empirical evidence, on the other. There was, as well, limited comparative analysis of populism across different parts of the developed world. Differences in political and electoral systems, in national circumstances, and in economic trajectories have frequently prevented this type of comparisons from being pursued. Lack of adequate data on interpersonal inequality has also represented a barrier for research on inequality in European regions. These problems have prevented shedding light on whether it was mostly unequal places or places left-behind that have driven the success of populist parties at the ballot box. Moreover, limited effort has been done to reconcile empirically racial/immigration anxieties and economic explanations behind the rise of populism, and particularly to study their joint effect in the rise of far-right populist parties.

In this article, we have addressed these issues. In line with theoretical expectations, the prolonged economic decline of regions is fuelling populism both in Europe and in the USA. The ‘revolt of the places that don’t matter’ has materialised in the ballot box (Rodríguez-Pose, 2018). In the USA, this is only the case in predominantly white-majority areas. More racially diverse, left-behind counties have not veered towards Donald Trump. A similar pattern is observed for the growth of interpersonal inequalities, suggesting that economic stagnation and rising inequalities polarise the vote between predominantly white areas and those that are more racially diverse. In Europe, the major factor explaining the support for far-right populist parties is the economic decline of regions, with most far-right populist voting concentrated in areas where GDP per capita growth has stagnated and immigration is relatively high.

While these results provide considerable food for thought and stress the need to revisit and analyse the link between interpersonal and territorial inequality and populism in greater depth, they should be taken with caution. Issues that would deserve greater attention than that afforded in an exploratory journal article are in evidence in our research. First, there is a need to improve data on interpersonal inequality. Most data on interpersonal inequality at a local level remain partial and often based on surveys of a limited number of households. Better and more consistent data are required to conduct a more in-depth analysis of what is a complex relationship. Second, both interpersonal and interterritorial inequality do not necessarily arise in wealth and monetary terms but may be more relevant in terms of deceived expectations and lack of opportunities (Goodwin and Heath, 2016b; Gidron and Hall, 2017; Rodríguez-Pose, 2018). Hence, other types of inequality, such as unequal access to educational and professional opportunities, could play a more important role in explaining the discontent at the ballot box. Third, more effort is required to understand how both cultural and economic factors affect different types of populist parties in Europe. It is also worth acknowledging that our lack of data on racial composition in European regions, as well as the fact that NUTS 3 regions in Europe and counties in the USA have different characteristics, imply that comparisons between the US and European cases must be interpreted with caution.

Taking these caveats in mind, our analysis suggests that grievances over an increasingly uneven distribution of resources, both within regions, in the form of rising local interpersonal inequalities, and especially between leading and left-behind regions, are good predictors of support for far-right populist parties. Addressing these concerns is essential to redress the dissatisfaction of citizens with the current economic and political system. The levelling up of increasingly unequal and left-behind regions would be a good starting point to do so.

Footnotes

Acknowledgements

The authors thank Harald Bathelt, the editor in charge, and four anonymous referees for their insightful and challenging comments to earlier versions of this article. They are also grateful to participants at seminar in Brussels, Dublin, London, Lyon, Pécs and Split for useful suggestions.

Funding

The LSE International Inequalities Institute's Cities, Jobs, and Economic Change Programme.

References

Annex 1. Growth in interpersonal inequalities and growth in GDP/income per capita growth in the USA and Europe

Annex 2. Variable descriptions and data sources