Abstract
What makes populist parties successful? What is the contribution of populist ideas to their appeal? How do they affect a voter’s choice? In this article, we treat populist orientation as just another dimension of spatial competition alongside economic and cultural dimensions, and assess its influence on vote choice. Using survey data from nine countries, we place voters and parties along the populist, economic, and cultural dimensions and estimate a spatial model of vote choice. We find that voters’ populist orientations are negatively correlated with their pro-market orientation and positively correlated with their cultural conservatism. We also find that proximity along the populist dimension has a tangible, albeit varying, weight in vote choice. This means that populist appeal can pull populist voters toward populist parties (and non-populist voters toward non-populist parties) even if they do not perfectly align with them on economic and cultural dimensions.
Introduction
The recent rise of populist parties has been commonly studied through the lens of specific policy issues and radical ideological positions. Such positions, arguably, meet the unmet policy demand at the opposite ends of the economic and cultural dimensions of the policy space (Dennison & Kriesi, 2023), and populism enables political parties to take on these positions (Kriesi, 2014; Mair, 2002). Nevertheless, such analyses reach their limit when attempting to explain the rise of populism in Europe. While populist parties do represent their constituencies well on certain issues (Backlund & Jungar, 2019), they do not perform better than non-populist parties (Kübler & Schäfer, 2022; Wuttke, 2019; Zhirnov, Thomeczek, Scotto Di Vettimo, et al., 2025). Moreover, the share of voters with radical economic and cultural positions is not large enough to explain why, for example, the newly founded Italian Five Start Movement won 25.5% of the votes in the 2013 elections, the Dutch Freedom Party (PVV) more than doubled its vote share (from 10.9% to 23.5%) and became the strongest party in 2023, and the Alternative for Germany (AfD) became the second strongest German party in the 2025 elections (rising from 10.3% to 20.8%). Can it be the case that there is more to the role of populism in electoral politics than creating a supply of parties with more extreme policy solutions? Does it possess an electoral appeal of its own and its own constituency?
Recent literature suggests that the move away from established politics and the increasing support for populists might be due to the proliferation of populist attitudes among voters and the demand for the representation of populist ideas (Akkerman et al., 2014; Marcos-Marne et al., 2023; Uscinski et al., 2021). However, the evidence of their effect is inconclusive. A few observational studies (Akkerman et al., 2014; Hawkins et al., 2020; Loew & Faas, 2019; Van Hauwaert & Van Kessel, 2018) find that voters with a populist outlook are more likely to vote for populist parties, even when other potential predictors of their appeal are included as control variables. Recent experimental studies (Castanho Silva et al., 2023; Dai & Kustov, 2024; Neuner & Wratil, 2022) arrive at different conclusions. They use conjoint experiments to conclude that making hypothetical candidates appear more populist does not have a more substantial effect on their appeal to populist voters than to non-populist voters. The main challenge to the observational results is that the complexity of the vote choice process makes it hard to appropriately account for all possible factors that could affect voting for populist parties beyond the voter’s agreement with their populist ideas. The main challenge to the null findings of the experimental studies is the difficulty of manipulating the populist component of a hypothetical candidate’s image: it is hard to make populist rhetoric authentic and sufficiently credible in experimental settings.
In this article, we join this discussion. We conduct an observational study and use statistical modelling to quantify the contribution of populist ideas to the appeal (or lack thereof) of populist and non-populist parties to populist and non-populist voters. To account for the complexity of vote choice and electoral markets, we resort to spatial models of voting (Adams, 2001; Davis & Hinich, 1966; Merrill & Grofman, 1999). Such models place political parties and voters in a common space according to their held or proclaimed preferences over public policies. Voters are assumed to base their choices among politicians on their expectations of the policies those politicians will pursue if elected. A vote results from a comparison among political options through the lens of the voter’s own views. This analysis situates each party within the context of other parties and breaks down each party’s and each voter’s ideological profile into its components.
To separate populist ideas from the economic and cultural dimensions of politics, we do not treat populism as a unique political phenomenon. Rather, we consider it as one of the tails of the populist dimension of political competition, similar to, but distinct from, its economic and cultural dimensions (we will refer to this dimension as populist orientation). The notion that populist orientation is a distinct dimension of politics is not entirely new to the literature. For instance, Uscinski et al. (2021) argue that there is an anti-establishment dimension in the American public opinion (mass opinion space). Marcos-Marne et al. (2020) treat populism as the third dimension in the cabinet formation process. The studies of the populist contagion – the use of populist rhetoric by mainstream parties in response to the rise of populist parties – implicitly assume the existence of a populist dimension (Kollberg, 2026; Krause et al., 2022; Rooduijn et al., 2012). That said, to our knowledge, no published research explicitly separates the contributions of congruence along economic, cultural, and populist dimensions of a political space to voting decisions. Without such an analysis, it is hard to tell whether the differential appeal of the populist ideas can help explain the support for populist parties. In this article, we take this step.
We take three main steps in our analysis. First, we locate parties and voters in a political space with populist, economic, and cultural dimensions. Second, we examine the pairwise correlation between voters’ positions along these dimensions. Finally, we estimate a spatial voting model to evaluate the weight of the distance between voters and parties along the populist dimension in voters’ vote intentions.
We rely on a cross-national survey that asked respondents to express their agreement with select policy- and populism-related statements, and an expert survey that captured political parties’ agreement with the same statements. Our cases include Czechia, England, France, Germany, Hungary, Italy, the Netherlands, Spain, and Sweden – a diverse set of European countries where populist parties have enjoyed varying degrees of success. We use this data and confirmatory factor analysis to generate country-specific political landscapes and locate both voters and parties on those landscapes. We find that populist orientation has a negative partial correlation with pro-market orientation and a positive partial correlation with cultural conservatism in all cases.
We use these landscapes to estimate a spatial model of vote choice. Voters are assumed to compare political parties based on the distances between them and the competing parties in the three-dimensional political space. We find that the voter-party distance along the populist dimension carries a discernible weight in vote choice. While this weight varies across political contexts, it is, on average, smaller than that of the programmatic dimensions (typically between 30% and 70% of the weight of the economic dimension and 10% and 60% of the weight of the cultural dimension). This means, for example, that the populist appeal of populist right parties can make them sufficiently attractive to center and center-left voters who hold populist attitudes. Likewise, populism can alienate rightist voters with more pluralist and pro-establishment views enough to make them vote for a centrist political party. In most cases, we also find evidence of complementarity between dimensions: voters tend to be more tolerant of the differences between them and parties if the parties are simultaneously more populist, more culturally conservative, and less pro-market than the voters and if they are simultaneously less populist, less conservative, and more pro-market than the voters.
The contribution of this article is threefold. First, to our knowledge, this is the first analysis to include populist orientation as a dimension in a spatial model. This allows us to examine parties’ populist orientations in the context of other parties’ and voters’ populist orientations and other elements of their ideological profiles. Second, we contribute evidence to the debate about the influence of demand-side populism on the vote choice. By incorporating the information about complete ideological profiles of all parties in the analysis and enabling voters to compare the complete ideological profiles of competing parties, we can better isolate the role of populist ideas and provide more robust evidence in support of the idea that populist ideas of political parties have a differential effect on voting among more or less populist voters. Third, we confirm the positive relationship between populist and pro-state economic orientations and between populist orientation and cultural conservatism at the voter level. The consistency of correlation patterns across country cases suggests that populism has socioeconomic and cultural roots. We do not find a similar consistency in the parameters of vote choice models. The significant variation in decision weights indicates that whether and how voter-level populism translates into political behaviors depends on the political context. The analysis of the features of that context that affect the activation of the populist dimension is an interesting avenue for future research, but it is outside of the scope of this article.
The remainder of this article is organized as follows. We start with a brief overview of the literature on populist attitudes and their relation to the populist vote. The following section details our thinking on populism as a dimension of the political space. We then present our data and analytical strategy and discuss the results. We conclude with a discussion of how our findings contribute to positioning populist orientation as a distinct political dimension in political analysis that is, in part, complementary to the cultural and economic axes.
Populism and Populist Attitudes
Populism attracted renewed interest in political science research in the early 2000s when observers were puzzled by the unexpected electoral successes of populist right parties in established Western European democracies. These parties did surprisingly well among the working class and the economically disadvantaged – the traditionally left-wing constituencies (Arzheimer & Carter, 2006; McGann & Kitschelt, 2005; Oesch, 2008). Something about populist right parties seemed to appeal to these new constituencies. These days, populist parties in European legislatures include not only the older populist right parties, such as Dutch PVV, but also populist left parties, such as Spanish Podemos, and more peculiar cases of populist parties that are neither radical left nor radical right, such as the Italian M5S and Czech ANO2011 (Kriesi & Pappas, 2016).
To establish a common denominator for identifying populist parties and to understand how populist parties differ from their mainstream counterparts, Mudde (2004) introduced the ideational definition of populism, which has since become widely accepted. According to this definition, populism is a “thin-centered ideology that considers society to be ultimately separated into two homogeneous and antagonistic groups, ‘the pure people’ versus the ‘corrupt elite’, and which argues that politics should be an expression of the volonté générale (general will) of the people” (Mudde, 2007, p. 23).
Much like most other ideologies, populism consists of a frame, which can be used to interpret the political environment, events, and relations, and a prescription for attitudes and behaviors (Busby et al., 2019; Denzau & North, 1994). Arguably, the frame is the key component of the ideational definition of populism. The opposites of the populist frame are those of pluralism and elitism (Mudde & Rovira Kaltwasser, 2013). Pluralists acknowledge the diversity of political preferences within society (and among elites) and view politics as a process of reconciling these differences. Elitism functionally separates the elites from the masses and views the former as better positioned and more apt to govern.
The opposition between populist views at one end and pluralism and elitism at the other is what makes it possible to measure the populist orientations of agents (and their messages). Individuals, parties, and organizations are populist to the extent that they endorse the populist frame and depart from pluralist and elitist views.
Early applications of this idea were on the supply side of politics, at the level of political parties and politicians. This research uses party manifestos, speeches, other types of party communication, and, more recently, expert surveys to capture the degree of populism of political parties and leaders (e.g., Düpont et al., 2022; Jagers & Walgrave, 2007; Meijers & Zaslove, 2020; Rooduijn & Pauwels, 2011). The ideational definition enables the populist orientation of a party to be treated as a continuum rather than a switch, thereby retaining information about varying degrees of populism in the ideas of populist and non-populist parties alike.
While ideologies can manifest in many forms, there is a long tradition in political science to focus foremost on the linguistic aspects of populism. In this sense, populist rhetoric is the most apparent expression of populist ideas (Aslanidis, 2015; Busby et al., 2019; Dai & Kustov, 2022; Gennaro et al., 2024; Licht et al., 2025). Such rhetoric interprets political reality in terms of the core populist ideas (Busby et al., 2019) and implicitly or explicitly communicates the degree of populism in the speaker’s worldview. Arguably, the use of such rhetoric can be part of parties’ strategies, much like the policy promises included in their programs.
The ideational definition of populism can also apply to voters’ views: just like voters can be described as more or less supportive of left-wing ideas, they can also be labeled more or less populist. Direct ideational measures of populism at the individual level were introduced in the early 2010s (Akkerman et al., 2014; Stanley, 2011) with the concept of populist attitudes. Populist attitudes can be defined as the extent to which individuals agree with the populist ideas and have been captured using specialized batteries of survey questions (Akkerman et al., 2014; Castanho Silva et al., 2020; Elchardus & Spruyt, 2014; Hawkins & Rovira Kaltwasser, 2018; Hobolt et al., 2016). The evidence from these studies shows that voter-level populism is quite widespread.
Arguably, both voters’ populist attitudes and party-level populism are manifestations of the same phenomenon, the same set of ideas. That said, studying the connection between demand-side and supply-side populism in electoral politics is not straightforward because voters’ and parties’ worldviews necessarily mix populism with other ideas. As a ‘thin-centered’ ideology, populism does not present a complete political worldview: it can be combined with more progressive and more conservative ideologies and needs a host ideology before it can be used in a political program. The reverse dependence is also true, as politicians rarely articulate some of the more radical ideas in host ideologies without referring to populist ideas (Roberts, 2018). That is why it is hard to separate populism from the host ideology in the appeal of a populist party to a voter who holds populist ideas.
Nevertheless, there have been a few attempts to link populist orientations on the demand and supply sides of political markets and study how the demand for populism meets its supply. The evidence produced by these studies is mixed. Some research shows that voters with populist attitudes are more likely to vote for populist parties (Akkerman et al., 2014; Hawkins et al., 2020; Jungkunz et al., 2021; Loew & Faas, 2019; Van Hauwaert & Van Kessel, 2018). These studies use observational survey data to compare the constituencies of populist and non-populist parties and/or model the probability of voting for a populist party. Akkerman et al. (2014) and Van Hauwaert and Van Kessel (2018) find that those who vote for populist parties are significantly more populist than those who vote for non-populist parties. Van Hauwaert and Van Kessel’s (2018) second analysis treats populist attitudes as predictors of voting for populist left and populist right parties. They find that populist attitudes are positively related to the populist party vote. Furthermore, holding populist attitudes can substitute for the lack of proximity between a populist party and a voter along the traditional dimensions. The effect of populist attitudes on the vote for a populist left party is stronger among right-wing and moderate voters than among left-wing voters. Hawkins et al. (2020) find a similar effect in their study of voting in Greece and Chile (their results are weaker in Chile).
These studies are not without limitations. It is hard to account for all possible determinants of voting for a populist left or right party in a reduced-form regression with control variables. Additionally, by treating party-level populism as a binary indicator, they lose information about the populism of non-populist parties and the different degrees of populism of populist parties 1 as well as a wide range of borderline cases not clearly classified as populist or non-populist (e.g., Forza Italia or Die Linke, see Rooduijn et al., 2023).
Contrary to the positive results in the above observational studies, recent conjoint experiments have failed to find a significant effect of populist attitudes on voting for populist candidates. The experiments conducted by Neuner and Wratil (2022) (in Germany), Castanho Silva et al. (2023) (in the US), Dai and Kustov (2024) (in the US) manipulate populism-related attributes of a candidate and measure the ensuing change in the candidate’s appeal to voters with higher and lower levels of populist attitudes. The differences between these groups of voters are not statistically significant. 2
One contribution of these studies is that they unpack the ideological profiles of populists; however, they might not be able to introduce sufficiently strong manipulation of the populist component of a hypothetical candidate’s profile, which could make that candidate credibly populist. Neuner and Wratil (2022) operationalize populism as having ‘populist priorities’, which include overthrowing incumbent elites, fighting corruption, and strengthening direct democracy. The study finds a differential effect of the most divisive priority – overthrowing the elites – on the choices of populist and non-populist respondents. The same pattern does not exist in cases where the hypothetical candidate prioritized anti-corruption efforts and the advocacy of direct democracy. Castanho Silva et al. (2023) add defending citizens’ interests and ending the abuse of power by parties to the list of populist priorities and find no statistically significant differences in the effects of advertising such priorities between populist and non-populist voters.
Dai and Kustov (2024) take an alternative approach to manipulating the hypothetical candidate’s populist orientation: rather than modifying a candidate’s declared priorities, they incorporate elements of people-centrism, anti-pluralism, and anti-establishment attitudes into candidates’ messages. They do not find statistically significant effects of these manipulations on the voting intent of populist and non-populist voters. While we find this operationalization more credible than the manipulation of candidates’ priorities, it might still not be strong enough to uncover the effect of populist ideas. Additionally, as in other studies, this study features a degree of mismatch between the conceptualization and measurement of supply- and demand-side populism.
In this project, we seek evidence that could reconcile these differences in findings and advance understanding of the link between the populist and programmatic dimensions of political competition. We make the operationalizations of supply- and demand-side populism more consistent by using the same model to generate voters’ and parties’ positions. This approach enables us to use populist orientation in a spatial model – the standard tool for studying how voters’ political views affect their choices – and to handle it the same way we would handle economic and cultural dimensions.
Programmatic and Populist Dimensions of Political Space
Political Space
Spatial models of elections have been used to understand how voters and parties interact on political markets since the 1950s (e.g., Adams, 2001; Aldrich & McKelvey, 1977; Davis & Hinich, 1966; Downs, 1957; Hinich & Munger, 1997; Ordeshook, 1986). Even though they may not be the best instruments for studying the psychology or social roots of voting, they help researchers disentangle the diverse influences of the political environment on vote choice. These models link a voter’s attitude towards a party to the party’s and the voter’s political views, place each party in the context of other parties, and separate different dimensions of political perspectives. They also allow us to consider the incentives of political parties to hold specific positions (regardless of whether they choose these positions or are ’born’ with them).
These models are built around voters’ and parties’ political views, which they represent as coordinates in a common space. A voter’s political view is complex. It includes their beliefs about how politics works and their roles in politics and society, their values and identities, and their preferences over what politicians should do and over political outcomes. These views may be organized by an ideology and political knowledge (Denzau & North, 1994; Jost et al., 2009), and in many cases, it is possible to reconstruct a narrative linking different components of a voter’s complete political view. That said, one of the long-standing results in public opinion research is that such constraints on individual belief systems are, on average, weak, and various components of a voter’s perspective rarely perfectly align (Converse, 1964).
Political parties also have political views and, more importantly for the analysis of electoral markets, a public image of those views. Such images are commonly at least as elaborate and complex as voters’ views. Unlike most voters, political parties spend time and effort communicating and defending their perspectives and policies. Parties’ ideologies, programs, rhetoric, and actions – all contribute to their images.
The spatial analysis of vote choice treats voters’ views and parties’ images as given, as exogenous to voters’ choices. This assumption may not be completely accurate: evidence shows that the elites can influence voters’ political views (Jost et al., 2009; Zaller, 1992). Political ideologies and discourse, which are shaped mainly by the elites, activists, and the media, provide mental maps that voters can adopt to formulate their views, connect ideas, and fill gaps in their views. That being said, the adoption of these maps is not automatic, and whether and how voters receive, choose, adapt, and deploy these mental maps varies, depending on various factors, including psychological and social factors and existing cognitive structures. The inertia in political views and their psychological and social roots is what allows us to think of voters’ views as given in the short term.
Spatial models represent voters’ views and parties’ images as points in a single multidimensional space. To serve its purpose, each dimension of such a space should represent a characteristic that (1) varies in levels/degrees, (2) can be used to describe both voters’ views and parties’ images, (3) cannot be reduced to the combination of positions along other dimensions, and (4) is relevant to voters’ satisfaction with political actions and outcomes.
There are different ways to define dimensions of the policy space that would satisfy these criteria. For example, they can be based on specific measurable traits of policy outcomes: the levels of taxation, the coverage of the healthcare system, the number and characteristics of immigrants, and the amount of red tape. Such outcomes will have the most immediate effects on a person’s life; however, the number of such items exceeds what a model can capture. Hence, it is more common to use ideological scales, which condense information about a person’s or party’s political beliefs and values and are predictive of their attitudes towards known and yet-to-be-known political outcomes (Enelow & Hinich, 1982; Hinich & Pollard, 1981).
Ideological positions are traditionally described as a single left-right dimension; however, this perspective can hardly give justice to the ideological diversity in the European party systems. Since the 1980s, European political spaces have been usually defined in terms of two ideological dimensions: economic and cultural (see, for example, Dalton, 2018; Inglehart, 1984). The economic dimension spans from the highest level of government involvement in economic activities on behalf of the public (on the left) to complete laissez-faire (on the right). It encompasses “pro-state versus pro-market orientations, disagreements over appropriate social welfare benefits, and the persisting issues of economic inequality” (Dalton, 2018, p. 29). The cultural dimension spans from the complete acceptance of new (post-materialist) values and minority lifestyles (on the left) to extreme cultural conservatism (on the right).
It is helpful to emphasize that economic and cultural dimensions do not apply exclusively to economic policies and cultural policies, respectively. Instead, any political decision can be characterized through the lens of both dimensions. For example, a law that regulates the access to and costs of health care can imply different degrees of government involvement in the activities of hospitals, pharmaceutical companies, and insurers, and various levels of redistribution (it can have different economic coordinates). This law may also exhibit varying degrees of bias towards majority norms and lifestyles, and be more or less inclusive (it can have different cultural coordinates).
Voters’ and parties’ coordinates along the economic and cultural dimensions are not uncorrelated: more pro-market voters are more likely to be culturally conservative. Benoit and Laver (2006) find consistent positive correlation between pro-market orientation and cultural conservatism in 41 of the 44 nations they examine. That said, the economic and cultural dimensions of the political space are distinct (Inglehart, 1984; Kitschelt, 1994; Roberts, 2018). They rely on different characteristics of policies, they are rooted in different kinds of values, and we cannot perfectly predict the degree of a voter’s cultural conservatism by knowing their position on the economic dimension.
Populist Dimension of the Political Space
We argue that populist orientation can be treated as an additional dimension of the political space distinct from its economic and cultural dimensions. Populist orientation is not perfectly aligned with either the cultural or economic dimension, just as the cultural and economic dimensions are also not perfectly correlated. Following the ideational definition, a political view can contain elements of the populist frame and the populist imperative blended with ideas from other ideologies. As Jagers and Walgrave (2007) put it, “populism, thinly defined, has no political colour; it is colourless and can be of the left and of the right.” Populist ideas are well compatible with both radical redistribution and an economy with minimal state presence, depending on how the ideology understands ‘the people’ and ‘elites’ in the populist frame and on the other elements of the underlying belief system (e.g., Pauwels, 2014). The same applies to the cultural dimension: an ideology with a populist component can be culturally liberal and culturally conservative (compare, for example, the populism of Podemos and Viktor Orban’s far-right populism). Populism does not require a host ideology of a particular type.
The ideational approach also helps us satisfy the other prerequisites of a dimension of a political space. First, this approach treats an agent’s populist orientation as a continuum: it is measured as the degree to which that individual or party agrees with populist ideas and distances themselves from pluralism and elitism (Hawkins & Rovira Kaltwasser, 2018). Second, this definition can be applied to both the supply and demand sides of electoral politics: both voters and political parties can be, to varying degrees, populist (Marcos-Marne et al., 2023; Van Hauwaert & Van Kessel, 2018). Evidence shows ample variation in populist attitudes among voters (Akkerman et al., 2014; Elchardus & Spruyt, 2014; Marcos-Marne et al., 2023; Spruyt et al., 2016), as well as the programs and rhetoric of political parties (Düpont et al., 2022; Meijers & Zaslove, 2020).
Populism can be relevant to voters’ expectations of political actions and outcomes, as well as their satisfaction with them. Populist orientations of a party are expected to manifest in its actions: more and less populist parties will differ in the ways they formulate new policies and interact with existing ones. Voters can expect a perfectly populist party to actively fight the entrenched elites, remove any opportunities for rent-seeking, never cooperate with mainstream politicians, maximize the involvement of ‘the people’ in the making and implementation of public policies, and follow its will in all policy-relevant choices.
Elitism and pluralism are commonly viewed as the ideological opposites of populism (Mudde & Rovira Kaltwasser, 2013). According to pluralist views, society includes individuals with different, sometimes opposing interests. There is no such concept as ‘the people’s will’ in this perspective; instead, the heterogeneity of the people is reflected in the political elite, which seeks compromises in governance. Elitism shares populism’s black-and-white thinking, but reverses the attribution: the people are not to be trusted and dangerous, while the elites have the natural right and competence to govern. In contrast to the elitism found in autocratic regimes, democratic elitism emphasizes the operational autonomy of the elites between elections and the professional nature of politics (Borchert, 2009).
Thus, the opposite end of the populist dimension (opposite to populism) is a combination of pluralism and democratic elitism. To link it to voters’ expectations about the beliefs and actions of a perfectly non-populist party, such a party can be expected to accept the diversity of society and the political class, and uphold the process of a representative democracy, including the operational autonomy of the elites. Such a party seeks cooperation with other parties, makes decisions through compromise, and relies on the technical elites and bureaucracy for advice and input in the formulation and implementation of policies.
Much like the economic and cultural dimensions, the populist continuum has implications for the content of policies; however, unlike the economic and cultural dimensions, it can also affect the preferred mode of politics for voters and parties. This makes this dimension somewhat special, but does not change the core logic of spatial analysis.
There are good reasons to expect voters’ positions along the populist dimension to be correlated with their positions along economic and cultural dimensions: they are linked through both the discourse and functional substructures. First, populist ideas are more compatible with some host ideologies than others. Huber et al. (2022) theorize the existence of collectivist (illiberal) and individualist (liberal) ideologies. The ideologies on the economic left (pro-redistribution and pro-state) are more collectivist than those on the economic right. The cultural right (conservative ideologies) is more collectivist than the cultural left. According to Huber et al. (2022), populism works better with collectivist ideologies: the collectivism of the economic left and the cultural right makes it easier to add content to the empty signifiers ‘the people’ and ‘the elite’. This idea is supported by Bakker et al. (2012) and Polk et al. (2017), who show that anti-elitist rhetoric is more prevalent among economic left-wing and culturally right-wing parties (see Tamaki & Jung, 2025, for a discussion at the voter level).
There is also evidence that positions along these dimensions share socioeconomic and cultural covariates. Voters with lower social status and those facing economic insecurity are more likely to hold populist attitudes (Abou-Chadi & Kurer, 2021; Gidron & Hall, 2017; Zhirnov et al., 2024): arguably, they resent the political and economic elites and view them as a reason behind their hardships. This effect is not confined to the lowest income strata; the “squeezed middle” class (Antonucci et al., 2017) is a known populist constituency. The same voters would also benefit from greater government involvement in economic activities and greater redistribution of wealth and economic security, both at the core of the left’s economic agenda. To put it simply, low status and insecurity make people resent the elites and support left economic policies. Accordingly, we expect voters’ pro-market orientation to be negatively correlated with their populist orientation.
Likewise, it has been argued that the support for populism has cultural roots. Spruyt et al. (2016) and Gidron and Hall (2019) argue that populism is, in part, driven by cultural anomie, limited social integration, and the rejection of prevailing social norms. These phenomena are more common in less urbanized areas, in more traditional environments, and in less educated segments of society, which also tend to be more conservative. Hence, we expect a positive correlation between voters’ populist orientation and cultural conservatism.
Vote Choice
Spatial models of vote choice view it as a comparison among parties’ positions – their coordinates in the assumed political space – relative to the chooser’s own position. Voters are more likely to vote for the parties that are closer to them in the political space than other parties. All else equal, they are more likely to vote for the parties that agree with them about the ideal role of government in the economy and the welfare state (have a closer position along the economic dimension) and about the tolerance for cultural changes (cultural dimension). We take it as a working hypothesis that they prefer political parties with more congruent beliefs about elites and the people in politics (populist dimension). In other words, voters with a more populist political outlook are more likely to choose a populist party over a non-populist party, and voters with a more pluralist and pro-elite outlook are more likely to select a non-populist party over a populist party.
The model aggregates the differences along the dimensions of the political space as weighted Euclidean distances between parties and voters (Davis & Hinich, 1966; Hinich & Munger, 1997):
The diagonal elements of this matrix are the weights assigned to proximity along specific dimensions and reflect their relative importance in a voter’s calculations. The higher the weight of the economic dimension, the more the voter cares about the congruence between the party’s perspective on redistribution and the welfare state and their own perspective, and the more likely they are to overlook differences in other components of the party’s platform and their own political outlook.
We take it as a working hypothesis that the populist dimension will have a positive weight in voters’ calculations: voters with a more populist outlook are more likely to vote for more populist parties, and voters with a more pluralist outlook are more likely to vote for more pluralist parties. The proximity of a party to the voter on the populist dimension increases its appeal and can compensate for incongruence on other dimensions. In other words, a more populist voter may care less about the economic and cultural implications of the policies a politician promises, as long as the politician shares their view that elites are malign and pledges to act in the people’s interest and against the elites. A less populist voter may do so if the politician shares their more pluralist perspective on politics, appreciates the role of the elites and bureaucracy, and promises to rely on science, experience, and compromise in their work.
We do not expect populist attitudes to translate into vote choice with the same weight across all countries. As Hawkins and Rovira Kaltwasser (2018) note, many people hold such attitudes but do not act on them. Such predispositions are activated (read: enter the vote choice calculations) in favorable political contexts, including the presence of political entrepreneurs willing to mobilize populist attitudes and their political orientations and the gravity of hardships that can be attributed to the actions or inaction of the incumbent elites (Hawkins et al., 2020; Hawkins & Rovira Kaltwasser, 2018).
Figure 1 illustrates what different estimates of the loss matrix can mean in spatial terms. It depicts a hypothetical political space defined by populist and economic dimensions, a voter (V), and the voter’s indifference curves – sets of political positions (i.e., political views) that the voter is indifferent among and is equally likely to support by voting. In panel a, the voter is equally likely to support parties P1, P2, and P4, which sit on the same ellipse, and is less likely to support party P3. These indifference curves are not circles: they are stretched to varying extents. How stretched they are along specific axes indicates how much weight voters attach to those axes. The observation that all ellipses are vertically stretched suggests that the voter attached a lower weight to the differences along the populist axis. A voter’s indifference curves in 2 dimensions. Note. The square marker represents the voter, and each contour represents that voter’s indifference curve: all the positions that give the voters the same utility.
The ellipses shown in panels a and b differ in whether their major axes are tilted relative to the space dimensions (they are tilted in panel b and parallel to the main dimensions in panel a). The loss matrices used to generate these ellipses differ in their off-diagonal elements: the major axes are parallel to the space dimensions only when the off-diagonal elements of the loss matrix are zero.
In substantive terms, when indifference curves are stretched diagonally, as in panel b, it means that the differences in a particular direction along one dimension may complement the differences in a specific direction along some other dimension and make them more acceptable to voters. Solid-lined ellipses in panel b are tilted to the left and represent the outlook of a hypothetical voter who, all else equal, favors more pro-state and, at the same time, more populist parties and more pro-market and, simultaneously, less populist parties over less populist and more pro-state parties and more populist and more pro-market parties. A party’s more populist orientation complements its more pro-state orientation, and a party’s more pluralist orientation complements its being more pro-market. Please remember that, in this model, parties’ positions are evaluated by each voter relative to their own position.
The off-diagonal elements of the loss matrix are often fixed to zero (the main exceptions are Stoetzer & Zittlau, 2015, 2020). However, this assumption makes vote choice models much less flexible; not flexible enough for our purposes. To obtain more general estimates of the role of populist attitudes in vote choice, we would need to allow for the possibility that parties’ more or less populist orientations complement the differences along the other dimensions, even if there were no theoretical reasons to expect such complementarities. That said, there are reasons to expect them. While a radical position does not necessitate populist rhetoric, populist rhetoric makes it easier for parties to persuade voters of the feasibility and merits of specific policies. According to Roberts (2018), political parties with radical left economic policies will use populist rhetoric to draw a line between them (the radical left) and the mainstream parties; those with culturally conservative positions will do so, too. It is plausible that higher levels of populism make it easier for voters to accept more pro-state positions than more pro-market positions, which suggests that more populist positions can complement more pro-state orientations (corresponding to the negative tilt in the indifference ellipses shown in Figure 1(b)). Populist frames also make it easier for voters to accept radical nativist and anti-immigration policies, which suggests that a party’s more populist position may complement its cultural conservatism.
There is another reason to expect complementarity between pro-state and populist orientations. The utility of economically expansive policies can depend on the practices used in their formulation and implementation. A decision to expand the regulation of a particular economic activity or industry is, according to the conventional understanding, an economically left-leaning action. At the same time, it may increase the opportunities for corruption and reduce the agency of the public, and as such, it may be less favored by voters who believe that the politicians implementing them are corrupt and will use this left-leaning shift to their advantage. To put it crudely, an ideal left government is fiscally and regulatory bigger than an ideal right government. The position along the populist dimension is predictive of how this bigger government will make specific spending and regulatory decisions, and whether those decisions will serve “the people’s will.”
Data and Empirical Strategy
Data and Measures
Our analysis relies on survey data from Czechia, France, Germany, Spain, Hungary, Italy, the Netherlands, Sweden, and England.3 These countries are quite diverse in their geographic locations in Europe, their electoral and party systems, and their welfare states. All of them, however, have experienced a recent rise in populism. We do not include cases outside Europe, which may limit the generalizability of our results to the blends of ideas specific to contemporary European politics, with a heavy share of parties mixing anti-immigration attitudes and Euroscepticism with populist ideas.
These data are part of a cross-national survey conducted in February-March 2023 as a follow-up to recent voting advice applications. The questionnaire includes statements about economic and socio-cultural issues and could be used to place respondents on the economic and cultural scales (Garzia et al., 2015). Additionally, it includes the battery of questions on populist attitudes developed and validated by Castanho Silva et al. (2020).
We also use expert surveys that assess the agreement of political parties with these statements. Most of these expert surveys were administered early in the campaign period. As a result, we have at our disposal information on the same items for both political parties and a sample of voters, and can apply the same model to recover the underlying factors. 4 This approach requires fewer assumptions to argue that voters’ and parties’ estimated positions are in the same space as when using different types of data sources for party-level and voter-level indicators.
Overview of Statements Used.
Notes. The populism battery includes the questions on populist attitudes developed and tested by Castanho Silva et al. (2020).
aThe response scale is reversed prior to the analysis.
bis replaced with ‡ in the analysis with the data from Spain, the Netherlands, and France due to the data availability at the party level and is completely excluded from the analysis with the data from Czechia.
We use confirmatory factor analysis without constraining the latent dimensions to be uncorrelated and estimate our models separately for each country sample. We include sampling weights based on respondents’ age, education, gender, region, and recalled vote in the preceding election. The measurement models fit the data well, with comparative fit index values ranging from 0.95 to 0.99. 5
We use these estimates to predict the latent variables for both survey respondents and political parties. Since populism is a combination of three components, and each component is necessary for an individual to be treated as populist (Jungkunz et al., 2021; Mohrenberg et al., 2021; Wuttke et al., 2020), we also generate a combined populism scale that treats each component as such. Specifically, we follow Jungkunz et al. (2021) and Wuttke et al. (2020) in measuring the overall populist orientation as the smallest value across the subdimensions of populism. Our measures are then standardized to have a mean of zero and a standard deviation of one.
Before analyzing vote choice, we validate our measures of party positions against a well-established continuous measure of populism, namely the POPPA populism measure (Meijers & Zaslove, 2020), and a well-established classification of political parties (PopuList project: Rooduijn et al., 2023). Both rely on expert opinions about the degree of populism – as defined by the ideational approach– in the rhetoric of political parties.
Figure 2 shows our estimates of party positions plotted against the POPPA estimates (Meijers & Zaslove, 2020) and indicates parties’ statuses in PopuList (Rooduijn et al., 2023). Without much surprise, our estimates are largely consistent with both the POPPA measure of populism and the PopuList classification: there is a positive correlation between the two continuous measures (at 0.66 in the full dataset and at 0.76 if we exclude Hungary), and almost all populist parties, as classified by PopuList, rank higher on the populist dimension than non-populist parties. Populist dimension and populist parties.
The main exception is the Hungarian Fidesz–KDNP alliance, which is classified as populist by both PopuList and POPPA, but has a very low populism score in our data. We suspect this is due to its long-standing incumbency in Hungary. Forza Italia, which was founded by a long-time prime minister Silvio Berlusconi, and ANO2011, which has also been in power, also score quite low on the populism scale. Even though this relationship between populism and incumbency can be viewed as a measurement problem and attributed to an undue emphasis that the anti-elitism questions placed on governing elites (Jungkunz et al., 2021), it is also relevant to the predictive power of populism for political practices and actions. Arguably, if a populist party is entrenched in government, it cannot credibly promise populist behaviors.
Modeling Vote Choice
To analyze voting, we estimate a series of random utility models, in which a voter’s utility from a particular party is driven by the weighted Euclidean distance between the party’s and the voter’s positions (see subsection “Vote Choice” in section “Programmatic and Populist Dimensions of Political Space”). We use different specifications for the loss matrix: 2- and 3-dimensional, with and without the off-diagonal elements fixed at zero.
Our model also includes behavioral components — voter-level variables (X
i
) and party-specific intercepts (γ
j
) (Adams, 2001). The former include age, gender, and education. The resulting utility of voter i from voting for party j can be formulated as follows:
As a robustness check, we further complicate this model by incorporating the possibility that voters discount the positions of more extreme political parties and/or use directional voting. Politicians are often constrained in the extent to which they can push policies and follow populist or non-populist political practices beyond the status quo. The promises of radical parties, thus, can rarely be taken at face value. According to Grofman’s (1985) model, voters discount the parties’ declared policy positions by taking a point between the declared policy positions and the status quo as the parties’ effective stances. To put it in Merrill and Grofman’s (1999) terms, rather than comparing political parties’ declared positions, voters compare their shadows. 6
Our dependent variable is intended vote choice. We exclude political parties with fewer than 10 voters in our dataset from the analysis. We use conditional logit to translate random utilities ui,j into predicted probabilities of voting for a specific political party and apply weighted maximum likelihood estimation, with the same sampling weights as those used at the measurement stage. Following Stoetzer and Zittlau’s (2015) approach, we replace Φ with a product of a lower triangular matrix and its transpose to constrain matrix Φ to be positive definite.
Results
Correlation Between Populist and Programmatic Orientations at the Voter Level
Figures 3 and 4 show the estimated political landscapes by country, and Figure 5 presents pairwise correlations among voters’ positions along the economic, cultural, and populist dimensions. We include both bivariate Pearson correlation coefficients and partial correlation coefficients in this display. Partial correlation coefficients are of primary interest here, as they reflect the association between pairs of dimensions after the potential confounding effect of the third dimension has been parsed out. Estimated landscapes. Estimated landscapes (continued). Bivariate and partial correlations between dimensions.
In all country cases, the partial correlation between voters’ pro-market and populist orientations is negative. Country-specific estimates range from −0.34 to −0.80, with a median of −0.58. Holding their cultural positions constant, more economically left (more pro-state) voters are also likely to have stronger populist attitudes.
The partial correlation between voters’ cultural conservatism and populist orientation is consistently positive, ranging from 0.27 to 0.70, with a median of 0.51. Holding their economic positions constant, more culturally conservative voters are likely to have stronger populist attitudes.
Populism in Vote Choice: Importance
AIC and Predictive Accuracy of Vote Choice Models.
According to this table, adding the populist dimension improves the model fit to the data in almost all cases, particularly if we focus on AIC. The accuracy metric is less supportive of this conclusion for Czechia, France, and Sweden (please note that accuracy may be misleading in multiclass classification problems with unbalanced categories). This suggests that the similarity between a voter and competing parties in the degree of their agreement with the populist worldview has explanatory power for voting, even after controlling for differences along the other two dimensions.
A similar conclusion can be drawn from the estimates of the loss matrix. Figure 6 summarizes the relative weights of the cultural and populist dimensions across our country cases and for different models. These weights are the square roots of the loss matrix’s diagonal elements. This matrix has been rescaled so that the economic dimension has a weight of 1. Each of these weights represents the utility from a hypothetical party one unit away from a voter along a specific dimension if the voter’s and party’s positions perfectly align along the other two dimensions. Weights of cultural and populist dimensions in vote choice. Note. Based on models 4 (with demographic variables but without discounting). The weights are the diagonal entries of the loss matrix rescaled so that the weight of the economic dimension is equal to one. The variables were scaled by country to have a standard deviation of 1 and a mean of zero before estimation. Whiskers indicate 95% confidence intervals.
The weight of the populist dimension is consistently below that of both the economic and cultural dimensions. Except for Sweden, where this weight is very close to zero (0.17), and the Netherlands (1.02), the populist dimension’s weight is between 0.35 and 0.70 of that of the economic dimension. These weights are not invariant to the rescaling of underlying positions. To anchor the scales of our dimensions, we have rescaled actors’ coordinates along specific dimensions to have a standard deviation of 1.
It appears that the cultural dimension has at least as strong an influence on vote choice as the economic dimension. It carries a significantly higher weight in voters’ decisions than the economic dimension in all countries except Spain and Hungary. The estimates range from 1.1 (in Spain, where it is not statistically significant) to 2.9 (in the Netherlands), with a median of 1.4.
Populism in Vote Choice: Complementarity
According to the model fit statistics (see Table 1), the models with an unconstrained loss matrix outperform those with the off-diagonal entries set to zero. In such models, a voter’s utility from voting for a party depends not only on how far the party is from that voter along specific dimensions but also on the direction of the party’s deviation from that voter’s position. The off-diagonal elements of a loss matrix capture the extent to which differences along specific dimensions and in specific directions can complement each other. 8
To give a substantive interpretation to the estimated complementarities, we calculate the utility that a voter accrues from hypothetical parties one unit away in different directions from her. Figure 7 shows these predictions as a heatmap, with darker colors corresponding to higher utilities (and higher propensity to vote for that party). Predicted values of voters’ utility from the parties 1 unit away in specific directions. Note. Based on model 4 with 3 dimensions. Darker colors represent higher utilities.
Our estimates reveal consistent complementarity between pro-market orientation and cultural conservatism. All else equal, voters tend to favor a more pro-market and more culturally conservative party and a more pro-state and more culturally liberal party over a more pro-market and more culturally liberal party and a more pro-state and more culturally conservative party. Cultural conservatism complements pro-market orientations. This result differs from the findings of Stoetzer and Zittlau (2015, 2020), who report complementarity between pro-state orientations and cultural conservatism (but, unlike us, force voters’ positions along the two dimensions to be uncorrelated at the measurement stage).
Our estimates in all cases except for Spain and Sweden show that parties’ more populist positions can complement their more pro-state orientations, and their more pluralist and pro-elite positions complement their pro-market orientations. This result is largely consistent with our expectations. All else equal, the voters in these countries prefer less pro-market and more populist parties (than the voters themselves) and more pro-market and less populist parties over other parties. Our estimates show the opposite in the Spanish case, which we speculate may be explained by the long tenure of the social-democratic party, PSOE, in government.
Most cases feature complementarity between the populist orientation and cultural conservatism. This means that, all else equal, voters prefer a more populist and culturally conservative party to a more populist party further to the left on the cultural dimension, and to a more conservative non-populist party. The exceptions are Spain, Sweden, Hungary, and Italy.
Our estimates of complementarity between the populist and programmatic dimensions show substantial variation across country cases, which may be surprising given how consistent our estimates of partial correlations between voters’ positions are. This observation, however, corroborates the activation argument (Hawkins et al., 2020): the regularity in the distributions of populist attitudes at the voter level can be attributed to their relatively stable socioeconomic and cultural roots, whereas the way populist attitudes translate into political actions is contingent on the political context.
Case Application: Data from Germany
To develop a geometric intuition for our estimates, we will take a closer look at a representative voter’s indifference curves in the German case. When the match between a voter and a party is imperfect, we can expect the differences between them to matter to varying degrees depending on the direction in which the party’s position departs from the voter’s stance. We find that if a party is more populist than the voter, it mitigates the effect of party-voter economic and cultural differences when the party is also more pro-state and culturally conservative, and increases their effect when the party is also more pro-market and culturally liberal.
Figure 8 shows the indifference curves drawn for a voter at the midpoint between AfD and CDU/CSU. This voter is economically and culturally right-leaning and is slightly populist. AfD is a populist right party, and CDU/CSU is a center-right, mainstream alliance. AfD and CDU/CSU are quite similar in their economic positions, but AfD is slightly more conservative and significantly more populist than CDU/CSU. Estimated voter indifference curves, Germany. Note. The square marker represents the voter, and the contour represents that voter’s indifference curve with all positions that give them the same utility as AfD and CDU/CSU. The position along the unmapped dimension is in parentheses.
Panel a shows the set of all points that give the voter the same utility as AfD and CDU/CSU as an ellipsoid in the three-dimensional space. The distance between the leftmost and rightmost points along a dimension reflects the relative weight of this dimension in voters’ decision-making: the wider the span, the lower its weight and the more willing voters are to accept the differences between them and parties along this dimension. Voters are more willing to tolerate differences along the economic dimension than along the cultural dimension. Judging by the span of the indifference curve along the populist dimension, the weight of the populist dimension is even lower than that of the economic dimension.
In the absence of complementarities between dimensions, the principal axes of the ellipsoid would be parallel to the dimensions of the political space. This is not the case. The main axis of the ellipsoid, along which it is most stretched, is the direction in which voters are most willing to tolerate incongruence in positions. This axis aligns with the unit vector [economic = −0.01, cultural = 0.12, populist = 0.99], which is almost parallel to the populist dimension but is tilted towards cultural conservatism and slightly away from pro-market orientations. This vector points toward more culturally conservative, more pro-state, and more populist parties. Because of the symmetry of the indifference surface, more culturally liberal, more pro-market, more pluralist parties enjoy the same advantage as more culturally conservative, more pro-state, more populist parties. The direction of the highest resistance is that of the unit vector [economic = 0.40, cultural = −0.91, populist = 0.12]: of more pro-market, culturally liberal, and slightly more populist parties.
Each of panels b, c, and d shows a planar cross-section of the indifference surface that is orthogonal to one of the dimensions and includes either AfD or CDU/CSU. The curve represents all the points that, all else equal, give the voter the same expected utility as the party under consideration and have the same value of the omitted dimension.
Cultural conservatism complements pro-market orientation, and cultural liberalism complements pro-state economic orientations, which can be observed in the shape of the indifference curve in panel b. As it illustrates, a party at [economic = 2.3, cultural = 2.0, populist = −2] carries the same utility as a party at [economic = 1.5, cultural = 0.3, populist = −2] even though the former is further from the voter. We can see that when a party is further along the economic dimension from a voter, that voter is more tolerant of the more conservative positions if the party is more pro-market and is more tolerant of the more liberal positions if the party is more supportive of state intervention in the economy.
The tilt of the ellipses shown in panels c and d similarly illustrates the complementarity between populist orientations and the programmatic dimensions. Such complementarities are also evident in panel b. All the points along the AfD curve are more populist than the voter (3.5 vs 0.8). Almost all of these positions are more culturally conservative than the voter: holding constant the level of populism, a more populist party is more attractive to a voter if it is more conservative than she is than if it were more progressive. That the centers of the curves are slightly to the left of the voter speaks to the complementarity between populist and pro-state orientations and between pluralist and pro-market orientations.
Conclusion
The notion that populism creates a distinct dimension of politics has been suggested by several authors (Marcos-Marne et al., 2020; Neuner & Wratil, 2022; Uscinski et al., 2021). However, we are not aware of any studies that have included the congruence between voters’ and parties’ populist orientations in a spatial vote choice model. By doing so, this study examines the demand-side populism in the context of the supply of populist options and the programmatic dimensions of political competition, and bridges it with the spatial voting literature. We treat populist orientation as just another dimension of the political space, placing voters and political parties within this space. We examine the distribution of their positions and estimate a spatial model of vote choice that includes both programmatic – economic and cultural – and populist dimensions on an equal footing.
As we examine the distribution of voter positions on the three dimensions of interest, we find a positive partial correlation between voters’ positions along the economic and cultural dimensions, a negative partial correlation between pro-market and populist orientations, and a positive partial correlation between cultural conservatism and populist orientation. The correlations between the populist and programmatic dimensions are never as strong as those between the economic and cultural dimensions, and the magnitude seems to vary more from country to country; however, they are consistently in the same direction. The patterns of correlation among voters’ positions along different dimensions are significantly more consistent across countries than the estimates of vote-choice models. We take it as suggestive evidence that there are demand-side roots of the positional implications of populism (the latter are discussed by Roberts, 2018). Arguably, these demand-side predispositions are driven by widespread declining socioeconomic conditions (Abou-Chadi & Kurer, 2021; Antonucci et al., 2017; Rodrik, 2017; Zhirnov et al., 2024) as well as by cultural divides (Norris & Inglehart, 2019; Spruyt et al., 2016). We do not have enough evidence to make causal claims, however.
Our analysis of vote choice suggests that populist orientation can indeed be treated as a dimension of the political space. We find that the proximity on the populist dimension can substitute for the lack of proximity on other dimensions. According to our three-dimensional spatial model, the weight a voter places on the distance between her and the party’s position along the populist dimension is smaller than that along the other dimension, but is not negligible (except in the Swedish case). Populist appeals will make a populist party more attractive to voters with populist attitudes, even if they do not entirely agree with its economic and cultural orientations. For example, an economically left or centrist voter with populist attitudes may vote for a populist right party due to its populist appeal. Similarly, avoiding populist appeals can make a party more attractive to voters with a more pluralist orientation (also see Kollberg, 2026). This evidence is consistent with the previous observational studies on populist attitudes (Akkerman et al., 2014; Hawkins et al., 2020; Van Hauwaert & Van Kessel, 2018), but runs counter to the prevailing null findings of conjoint experiments (Castanho Silva et al., 2023; Dai & Kustov, 2024; Neuner & Wratil, 2022). We suspect that part of this difference is due to the difficulty of manipulating the degree of populism in a hypothetical candidate’s image to the extent necessary to identify the relatively weak effect of populist orientation, and the other part is due to differences in the operationalization of populism.
In contrast to earlier research, which theorizes that the main appeal of populist parties may lie in their ability to take on underrepresented positions on cultural issues, immigration, income gaps, and radical taxation (Kriesi, 2014; Mair, 2002), our result suggests that the growing support for populist parties can be due to the spread of populist ideas proper, and part of the appeal of mainstream parties can come from the fear toward populist parties among the voters with more pluralist and pro-establishment views. These considerations may loosen the link between voters’ and their representatives’ positions along economic and cultural dimensions.
When we theorize about the populist dimension, we consider it as one of the characteristics of a party’s image that suggests to voters how the party would act if elected. Much like their positions along the economic and cultural dimensions, their populist orientation can be mapped to specific variations in the content of policy proposals. In addition to policy implications, populist orientation affects parties’ and voters’ outlooks on procedural (including constitutional) matters: cooperation with other elites and protection of minorities. This makes it somewhat different from the programmatic dimensions.
Accordingly, if there is a heightened level of populism in society, we should expect to see more elected politicians who prefer conflict over cooperation and may be willing to disregard minority rights. The latter may have a limited effect on political outcomes, as modern democracies have constitutional safeguards against illiberal politics, such as constitutional guarantees of human rights and individual liberties, as well as independent courts. Populist politicians rarely have enough legislative clout to bypass these safeguards; however, this is a distinct possibility.
In addition to the discernible weight of the populist dimension, we also find complementarity between economic and populist orientations. In most cases, voters are more tolerant of parties that are more populist and favor state-led economic intervention, and of those that are more pluralist and pro-market than the others. As suggested earlier, this can be driven by a preference for a larger government only if it is more committed to the will of the people. It could also be due to the compatibility of populist discourse with more collectivist ideologies (this resonates well with the phenomenon of welfare populism, see, for example de Koster et al., 2012). We also find examples of complementarity between cultural conservatism and populism.
It is helpful to emphasize the significant variation in how voters make decisions across our cases. Despite quite similar estimates of the relative weight of different dimensions in vote choice and comparable estimates of complementarity between economic and cultural dimensions, there is significant variation in the estimated complementarities between the programmatic and populist dimensions. This result agrees with Rooduijn’s (2014) conclusion that the strains of populism found in different countries are quite unique. Part of this variation is due to the differences in the historical configurations of political parties in the system. Some countries only have a dominant populist far-right party, while others have relevant populist parties from different political camps, or even multiple populist parties from the same camp. There is also variation in how populist attitudes are activated into political action (Hawkins et al., 2020; Hawkins & Rovira Kaltwasser, 2018). Future research will have to explore the determinants of such variation.
One limitation of our approach is that standard spatial models cannot easily capture the variation in the salience that different voters attach to different dimensions. For example, voters may discount the differences along the other dimensions as long as the parties offer sufficient congruence along the dimension of the highest salience to them. Kirkizh et al. (2022) show that the supporters of radical right parties do not mind political parties’ offering policies of different economic orientations as long as they support restrictive anti-immigration policies. It is also possible, as shown by Van Hauwaert and Van Kessel (2018), that more populist voters place a different value on programmatic dimensions than less populist voters. To test these possibilities, we estimate versions of our models with separate loss matrices for more and less populist voters. We do not find that the dimension weights differ systematically between these two groups of voters (see section F of the Supplemental Information); however, further research is needed to confirm this result.
Another limitation is that we cannot take into account the role of the elites in shaping voters’ political views. That said, the stability of the patterns of correlation across dimensions in different countries and voting weights suggests that there are exogenous demand-side factors that have a significant structuring effect on the voters’ policy preferences and populist orientations.
We should also acknowledge that we do not examine cases outside Europe; thus, our results may not generalize well to Latin America (or other regions of the world), where populism appears in different forms than in Europe. European democracies have fewer cases of a mix of left-wing economic ideas and populism than Latin America. The Spanish case, where the populist left Podemos had a strong effect on political discourse, hints at what we could observe if we had data on Latin American democracies. More research is needed to confirm this idea.
The results of this study have broader implications for political science research, despite its limitations. First, we demonstrate that populist orientation can be studied as a dimension of political competition, rather than merely a predictor of policy positions along economic and cultural dimensions, or a phenomenon that exists outside electoral competition. Spatial models of elections may benefit from incorporating populist orientation as a separate dimension of party competition. Second, our results suggest that taking on more populist stances is a plausible, albeit not necessarily effective, strategy for mainstream political parties in response to the threat of rising populism, consistent with the populist Zeitgeist thesis (Mudde, 2004). While there is mixed evidence on whether mainstream parties have adopted such strategies (Breyer, 2022; Rooduijn et al., 2012), we expect more research on the strategic implications in the near future.
Supplemental Material
Supplemental Material - A Populist Axis? Analyzing Connections Between Populist, Economic, and Cultural Dimensions of Political Space
Supplemental Material for A Populist Axis? Analyzing Connections Between Populist, Economic, and Cultural Dimensions of Political Space by Andrei Zhirnov, Jan Philipp Thomeczek, Lorenza Antonucci, Michele Scotto Di Vettimo, Roan Buma, Vladimir Cristea, Norbert Kersting, André Krouwel, Sven Lange, López Ortega Alberto, Jack Thompson, Roberta Di Stefano, and Ruggeri Francesco in Comparative Political Studies.
Footnotes
Acknowledgments
We are grateful to Susan Banducci, Cecilia Meneghini, as well as Rebecca Kittel, Roman Chytilek, and other participants of the ECRP workshop “Through Thick and Thin: Bridging Innovative Conceptual and Methodological Approaches in Studying Programmatic and Discursive Party Competition” (2023) for their helpful comments, questions, and suggestions. All remaining errors are our own.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: this work was supported by the Volkswagen Foundation [PRECEDE project, ref. 96999].
Ethical Considerations
The data used in this paper comes from a cross-national survey conducted by a Dutch survey company Kieskompas B.V. in Czechia, the United Kingdom, France, Germany, Hungary, Italy, the Netherlands, Spain, and Sweden. The project went through the ethics review self-check provided by the FSW Research Ethics Review Committee (RERC) at Vrije Universiteit Amsterdam (ref.: 2023-9-26–801). Informed consent was obtained from all participants before proceeding to the survey. All participation was anonymous and was not compensated. No deception was used in the surveys.
ORCID iDs
Data Availability Statement
Replication materials and code can be found at Zhirnov, Thomeczek, Antonucci, et al. (2025).
Supplemental Material
Supplemental material for this article is available online.