Populism and Candidate Support in the US: The Effects of “Thin” and “Host” Ideology

Abstract

Much of the contemporary literature on populism focuses on its status as a “thin” ideology comprising three key components: people-centrism, anti-elitism, and anti-pluralism. Populist politicians pair this “thin” ideology with extreme positions on policy issues such as immigration or taxation (referred to as “host” or “thick” ideologies). A recent study using German samples leveraged conjoint experiments to disentangle the effects of these appeals on vote choice. The results not only showed that extreme host-ideological positions mattered more than so-called “thin” populist appeals, but also that effects of populist appeals were nearly identical among populist and non-populist voters. Our replication in the US context reaffirms both the importance of host-ideological positions and the lack of heterogeneous effects by voters’ “thin” populist attitudes. Furthermore, by uncovering some divergence from the German case (e.g. anti-elite appeals trumping people-centric appeals), we highlight the need to experimentally examine the effects of populism’s constituent components across contexts.

Introduction

As populists score electoral victories across the globe, from India to the US to Hungary, the extant political science literature has begun to converge on a common understanding of populism. The so-called Ideational approach (Hawkins et al., 2018) posits that populism is a “thin-centered” ideology (Mudde, 2004), meaning that it comprises a few basic tenets but can be combined with a range of different “host” (or “thick”) ideologies from across the political spectrum. The three basic components of populism are: a) people-centrism: a romanticized view of common people and a belief that implementing the people’s will is the only goal of democratic politics; b) anti-elitism: perceiving existing political or economic elites as an evil group keen on oppressing common people; and c) anti-pluralism: the belief that politics is a struggle between good (the people) and evil (the elites), in which there are no other meaningful cleavages (e.g. Aslanidis, 2016; Canovan, 1981; Canovan, 1999; Hawkins and Kaltwasser, 2017).

Empirical research relying on observational data has consistently found that populist attitudes, operationalized based on this “thin” ideology definition, predict support for populist parties and politicians, alongside voters’ particular, extreme issue positions (e.g. Akkerman, Mudde and Zaslove, 2014; Castanho Silva et al., 2020; Hawkins, Kaltwasser and Andreadis, 2020; Hieda, Zenkyo and Nishikawa, 2019; Marcos-Marne, Plaza-Colodro and Freyburg, 2020; Van Hauwaert and Van Kessel, 2018). However, little is known about which components of this thin populist ideology are driving this support and how these factors combine with the host ideologies – such as anti-immigration or pro-redistribution stances – that populist politicians adopt. Indeed, Art (2020) recently argued that the emphasis on thin ideology is obscuring the role of nativism as the defining feature of populism. More generally, Hunger and Paxton (2021) argue that researchers often conflate populism with host ideology and suggest that scholars clarify whether they are examining the role of thin ideology (e.g. people-centric or anti-elite appeals) or of substantive issue positions (i.e. the host ideology).

Some studies have sought to disentangle the role of host ideologies and thin populist ideology as well as their constitutive components. Research on support for radical right parties has long highlighted the importance of left-right ideology and issue attitudes, in particular anti-immigrant sentiment (i.e. a key host ideology), while finding little evidence for explanations related to thin populist ideology, such as voters’ protest motives (e.g. Van der Brug, Fennema, and Tillie, 2000; Van der Brug, Fennema, and Tillie, 2005). More generally, Loew and Faas (2019) show that the much touted trifecta of people-centrism, anti-elitism, and anti-pluralism only partially explains support, and that issue positions have at least as large an impact, while Bakker, Schumacher and Rooduijn (2021) show that both anti-establishment and host ideology messages matter and that different personality traits moderate these relationships. In a different vein, Busby, Gubler, Hawkins (2019) show that framing populist rhetoric using dispositional blame as opposed to situational blame increases support for populists, and that this effect is moderated by populist attitudes.

Neuner and Wratil (2022) were among the first to approach this issue experimentally using conjoint experiments fielded on representative samples in Germany. Their results suggest that while some thin people-centric political priorities appeal to voters, others such as anti-elite appeals do not. Importantly, the authors found no evidence for differential effects of these appeals by respondents’ populist attitudes. Moreover, positions on host ideology issues such as redistribution and immigration exerted greater influence on populist respondents’ vote choice. ¹ However, as Neuner and Wratil (2022, p. 563) themselves note, the German case might not be generalizable, and thus the design should be replicated across a variety of contexts. We take up this call and replicate their design in a different context: the US in 2019, which featured a populist president in office (Hawkins and Littvay, 2019), a different party and electoral system as well as different dimensions of political competition. Our findings suggest that Neuner and Wratil’s (2022) main results hold: the most effective populist appeal is a host ideology appeal and there are no heterogeneous effects by respondents’ thin populist attitudes. In contrast to the original study, however, we find limited effects of thin people-centric appeals but a strong effect of an anti-elite appeal criticizing political parties, suggesting that the power of various thin priorities differs across contexts. These results confirm the need for a reappraisal of research on populist attitudes, which seems to overemphasize the role of thin ideology, while overlooking the actual issue positions that attract voters to populists.

The Original Study

Neuner and Wratil (2022) leverage conjoint experiments to evaluate which components of candidates’ populist ideology garner electoral support. Their paired conjoint design presents respondents with five pairs of fictitious candidates who exhibit a range of different attributes, and in each of the five rounds respondents have to choose which of the two candidates they would rather vote for in a “Bundestag” election. Candidates’ populism is operationalized with six attributes. First, candidates are described as having a first and second political priority, which can include populist valence issues such as “Overthrow the political elite” or “Strengthen direct democracy,” but also non-populist priorities such as “Fight crime” or “Promote economic growth.” Second, candidates’ descriptions include their positions on four host ideology issues on which populist parties often embrace a particular position: a) refugees (with levels ranging from accepting many to deporting many); b) taxation of the rich (from much lower to much higher taxes); c) position on the EU (from more integration to leaving the EU); and d) trade and globalization (from much more to much less). Positions advocating for the deportation of refugees are usually linked to right-wing populism, whereas positions advocating for the taxation of the rich are associated with left-wing populism, while anti-globalization as well as anti-EU stances unite both left- and right-wing populists. ² The authors also measure respondents’ populist attitudes with a range of items extending the commonly used populist attitudes scale developed by Akkerman, Mudde and Zaslove (2014).

The original study finds that, among the general public, two ideological positions often held by populists increase the probability of a candidate being chosen (deporting many refugees and taxing the rich), while two issue positions often held by populists decrease that probability (Euroscepticism and being anti-globalization). Regarding thin populist appeals, the people-centric “Strengthen direct democracy” and “Defend citizens’ interests” increase support, while the anti-elite “Overthrow the political elite” is a clear electoral liability and “End the abuse of power by the parties” has no effect. When breaking the results down by voters’ pretreatment thin populist attitudes, Neuner and Wratil (2022, p. 567) find that these attitudes have little impact: there are “virtually no differences in the appeal of populist priorities by populist attitudes. Thin populists are not significantly more swayed by thin populist appeals than non-populists or thick populists.”

Adapting the Design to the American Context

Replicating Neuner and Wratil (2022) in the US context is important. First, while the US has a history of populist movements, it traditionally lacks major parties commonly classified as “populist” (such as the AfD or other right-wing parties in Europe), thus providing a party system context that differs markedly from the German and many European cases. That said, the recent rise of Donald Trump and Bernie Sanders as well as the Tea Party and Occupy Wall Street movements have reignited debates about populism in the US (Oliver and Rahn, 2016; Bonikowski, 2016). Though note, Art (2020) points out that it is questionable whether the defining feature of prominent US populists such as George Wallace or Donald Trump is (thin) populism or nativism (i.e. a host ideology), thus highlighting the need to disentangle the two. Second, empirical research on the impact of individual-level populist attitudes on vote choice has been largely confined to Europe and Latin America (e.g. Andreadis et al., 2018; Van Hauwaert and Van Kessel 2018).

To closely replicate the original study, we framed the experiment in terms of a choice between two candidates in a Congressional election (as the counterpart to the “Bundestag”), without providing party labels (see Online Appendix C and discussion in the conclusion). ³ However, we adapted the design slightly to ensure that attributes reflect salient domains of US party competition. A side-by-side comparison of the original German conjoint and our US conjoint is presented in Online Appendix Table A.2. For candidates’ priorities, we deleted the context-specific priorities “Lead Germany out of the crisis” and “Create a social Europe” but included “End the abuse of power by the parties,” a less charged anti-elite appeal that Neuner and Wratil (2022) used in their amended design. ⁴ For candidates’ positions, we changed the immigration attribute from being about “refugees” to “legal immigrants,” which better reflected the US discourse at the time. Lastly, we replaced the EU cooperation issue with the issue of military intervention, which constitutes a similarly isolationist foreign policy perspective that makes sense in the US context. As isolationism and non-interventionism constitute populist positions in Jacksonian, Tea Party, and Trumpian populist narratives (see e.g. Mead, 2011; Lacatus, 2020; Kazin, 2016), we consider advocating for “much less military intervention” to represent the populist position.

A central concern in Neuner and Wratil (2022) was examining heterogeneous effects by both thin populist attitudes and positions on host ideology issues commonly associated with populism (referred to as “thick” populist attitudes in the original study). Briefly, to capture thin populist attitudes, we construct a latent variable model based on items from the Akkerman, Mudde and Zaslove (2014) populist attitudes scale as well as additional items suggested by Castanho Silva et al. (2020) and classify respondents below the mean as “non-populists” and those above the mean as “populists.” ⁵ Following Neuner and Wratil (2022), we also attempted to measure populist host ideology attitudes (i.e. views toward legal immigration, globalization, taxation of the rich, etc.). However, contrary to the German case, these variables failed to load onto a single factor, suggesting that these views are less consistently correlated in the US. Consequently, given the importance of partisanship in the US, we disaggregate the results by thin populist attitudes as well as partisanship. Online Appendix E provides comprehensive information on the measurement of populist attitudes.

Data

We fielded our experiment on a broad national sample of US adults (n = 3,024) provided by Lucid with quotas for gender, age, ethnicity, and region between July 30 and August 4, 2019 (Castanho Silva, Neuner and Wratil, 2022). Estimates from Lucid samples have been shown to converge to those from US high-quality probability samples and to outperform convenience samples such as Amazon’s MTurk (Coppock and McClellan, 2019). Only half of the sample was randomly allocated to participate in the conjoint discussed here, and thus the resulting sample size for this experiment is 1,505. Including leaners, 39% of respondents identify as Republicans, 48% as Democrats, and 13% as Independents, which is in line with estimates from probability samples (e.g. ANES) (see also Online Appendix A).

Results

Figure 1 shows the results of our experiment for the full sample. ⁶ The estimates are marginal means (MMs) and the dashed line denotes the 0.5 probability of a candidate being chosen, meaning that estimates to the left indicate that the attribute level has a negative effect on vote choice, while estimates to the right signify a positive effect. ⁷ Among the thin populist priorities, we find that some lead to increased support among the general public: “Fight political corruption,” “End the abuse of power by the parties,” and “Defend citizens’ interests.” However, “Overthrow the political elite” and “Strengthen direct democracy” have negative albeit nonsignificant effects on the probability of the candidate being chosen. In contrast to Neuner and Wratil (2022), our results suggest a stronger appeal of “soft” anti-elitism in the US. While Germans were very clearly repelled by the idea of overthrowing the political elite, this effect is much smaller and marginally not significant in the US. In addition, appealing to “End the abuse of power by the parties” is a significant electoral asset for candidates in the US – but had no effect on candidate choice in Germany. In turn, people-centric appeals had some consistent, positive effects in Germany, whereas in the US, “Defend citizens’ interests” is a small asset but “Strengthen direct democracy” presents a liability.

Figure 1 Marginal means of attribute levels. Notes: Marginal means; 95% confidence intervals as horizontal bars; dashed line indicates Pr(Y = 1) = 0.5.

Regarding populist host ideology positions, we find that the populist call for higher taxes on the rich also resonates in the US. Like in Germany, populist anti-globalization stances do not garner support. The effects of anti-immigrant positions are similar in structure but clearly muted in the US. We also find that the populist position of “less military intervention” significantly increases support. In Figure 2, we zoom in on those attribute levels representing populist priorities and positions and examine whether their effects vary by respondents’ populist attitudes. In line with the German results, we find no evidence for effect heterogeneity by thin populist attitudes, as both populist and non-populist US respondents react to candidates’ populist positions on host ideology issues (top panel) as well as thin populist priorities (bottom panel) in the same manner. This is particularly noteworthy for the thin populist priorities (e.g. “End the abuse of power by the parties”), as populist attitudes scales were developed to capture voters’ preferences on precisely such appeals. In fact, none of the differences in MMs between populist and non-populist respondents are close to being statistically significant. ⁸

Figure 2 Marginal means of populist positions (top) and priorities (bottom) by respondents’ populist attitudes. Notes: Marginal means; 95% confidence intervals as horizontal bars; dashed line indicates Pr(Y = 1) = 0.5.

Next, we disaggregate the results of the populist attribute levels by respondents’ partisanship in Figure 3 given its importance in the US context (see also Online Appendix G). First, compared to Republicans, Democrats prefer candidates who support much higher taxes on the rich and reject those who want to greatly decrease the number of legal immigrants, reflecting both parties’ stances on these issues. Second, regarding thin populist priorities, the results corroborate the finding that the effect of thin populist appeals is largely homogeneous across individuals, as differences between the groups are small, except for the “Defend citizens’ interests” position, which significantly increases support among Republicans but not Democrats or Independents. There are no other priorities for which we find a statistically significant difference between Republicans and Democrats.

Figure 3 Marginal means of populist positions (top) and priorities (bottom) by respondents’ partisanship. Notes: Marginal means; 95% confidence intervals as horizontal bars; dashed line indicates Pr(Y = 1) = 0.5.

Conclusion

Scholars who find that populist attitudes predict support for populist parties in observational studies usually interpret this as evidence that voters are drawn to those parties because of the thin ideological components of the parties’ populist discourse. Neuner and Wratil (2022) use conjoint experiments and show that this is not the case in Germany, and we substantively replicate this finding in the US. First, we find that a populist issue appeal (taxing the rich) has the largest effect on candidate support. Second, while some thin populist priorities such as fighting corruption and combating party elites increase support, they do so universally rather than being moderated by respondents’ populist attitudes. This raises doubts about whether “populist voters” are uniquely mobilized by populist appeals and/or whether current survey scales measuring thin populist attitudes truly capture a disposition that influences populist voting.

One caveat worth noting is that we did not include party labels in the experiment. We did so primarily to closely mirror the original study. Moreover, as noted above, there was no moderation by thin populist attitudes in the original study. As party labels tend to diminish or crowd out other effects on candidate choice (Kirkland and Coppock, 2018), we wanted to ensure that any replication of a null finding on populist attitudes in the US could not simply be due to the inclusion of party labels. That said, we conducted exploratory analyses to probe whether the inclusion of party labels would possibly change our results and whether populist appeals would be more persuasive if tested in a party primary setting, where thin ideological appeals may be more influential (see Online Appendices I and J). Despite low power, these analyses provide suggestive evidence that when respondents can infer the candidate’s party label from their host ideology positions, this does not seem to alter the effect of other attributes. Moreover, they provide no evidence that the effect of thin populist attitudes would be heightened in contests where candidates with similar host ideology are pitted against one another (i.e. a primary setting). Future research should more directly test how populist appeals fare in experimental settings where party labels are varied or where respondents are asked to choose between candidates from the same party.

In conclusion, our findings echo those of Neuner and Wratil (2022) in suggesting that the role of thin populist attitudes may be exaggerated, and that the typical populist positions on host ideology issues should play a more central role in our explanations of the global rise of populism. Finally, by uncovering some divergence from the German case in the precise components of populism that matter to voters (e.g. anti-elite as opposed to people-centric appeals), we further highlight the need to disaggregate populism into its constituent parts and to experimentally examine the effects of these various components across contexts.