The rhetorical contestation of populism in four European parliaments (2010–2020)

Abstract Studies of the discursive aspect of populism have tended to focus on the rhetorical manifestations of populist discourses, namely how populist politicians ‘talk politics’. However, scant attention has been devoted to how politicians rhetorically characterise the concept of populism itself. This article investigates a corpus (N = 4.835) of parliamentary debates in Germany, Italy, Spain and the United Kingdom, to assess the different definitions of populism used by politicians and explain any variations found. Results show that the ways that politicians describe populism vary cross-nationally in terms of topics, adjectives and targets. A more positive attitude towards populism is likely for politicians who express anti-elitism and opposition to multiculturalism and corruption, regardless of the topic under discussion. Moreover, the effect of political ideology on politicians’ attitudes towards populism is more pronounced when economic topics are under discussion. Discussing these findings, this study offers interesting implications for the literature on populism, parliamentary affairs, party politics and text analysis. Supplemental data for this article can be accessed online at: https://doi.org/10.1080/01402382.2021.2013655 .

on how to identify, measure and characterise populism (Moffitt and Tormey 2014;van Kessel 2014;Wuttke et al. 2020). Many works have focussed on the discursive aspect of populism, but although the literature has often investigated the rhetorical manifestations of populist discourses -how populist politicians 'talk politics' (Caiani and Della Porta 2011;Elmgren 2018;Hawkins 2009;Stavrakakis 2017) -scant attention has been devoted to how populist and non-populist politicians characterise populism. Hence, this article tackles these crucial, yet mostly unaddressed, questions: How do politicians define populism? Why do politicians characterise populism in different ways?
Such an investigation provides important contributions to the academic literature on populism, party politics, parliamentary studies and rhetorical analysis. First, this article contributes to the understanding of populism and its complex nature, offering an overview of politicians' interpretations, which is different from previous studies that have mostly focussed on academic conceptualizations. Second, given the popularity of 'populism' in the public debate, it is crucial to understand how and under what conditions politicians construe and employ the term. In particular, there is a need to test whether the dimensions (e.g. ideology, topic, strategic incentives) that influence the way MPs characterise populism in the parliamentary arena reflect the depiction of populism that is prominent in academic and public debates.
Moreover, the analysis of party-level factors in explaining different attitudes towards populism can contribute to the recent debate on the evolution of political parties' ideologies and behaviour (Ignazi 2017;Jackson and Jolly 2021;Wuttke 2020), assessing how they position themselves on the debate on populism. Finally, the coding of a corpus of parliamentary debates on a crucial contemporary issue may provide an additional key contribution, as our dataset could be employed as a training set for quantitative analysis on larger corpora, as well as offering guidance for the analysis of multilingual corpora. 1 In particular, we investigate in the period 2010-2020 all paragraphs (N = 4.835) in which MPs from Germany, Italy, Spain and the United Kingdom mention the keyword 'populism' during their debates in the parliamentary arena. As shown in more detail below, we chose these countries in order to explore variation in terms of electoral success and typologies of populist parties, other than the topics that revolve around populism. Our results show that parties' higher scores on the core dimension of the populist discourse -anti-elitism -determine a more positive characterisation of populism. This is an insightful result, as previous research has shown that populist parties also often tend to characterise populism in negative terms, likely following the general stigmatisation of the term (e.g. Casiraghi 2021).
Second, multiculturalism emerges as a key driver of MPs' rhetorical behaviour: harsher stances on multicultural policies increase the positive interpretation of populism, showing also the relevance of the connection between the two political issues in the contemporary discussion. Finally, the impact of political ideology is more complex and varies depending on the topic of the debate. However, our results suggest that MPs on the right are more prone to characterise populism in negative terms, a surprising result that contradicts the 'populist-hype' thesis that is prevalent in academic and public debates. In addition, such a finding provides a new perspective when thinking about the 'populist contagion' that should have characterised especially right-wing parties across Europe (e.g. Schwörer 2021).
The article is structured as follows. The next section presents our theoretical framework and the hypotheses. Section three presents the data collection process and thoroughly explains the text-analysis we have performed. Finally, section four discusses our results, with the conclusion summing up our findings and describing some implications.

From populist discourses to discourses on populism
As scholars have noted, after decades of debates about the definition of populism and of populist discourses, there is the need to investigate more thoroughly the characteristics and determinants of discourses on populism to further our understanding not only of populism, but also of the democratic political competition (De Cleen et al. 2018;Kefford et al. 2021;Rooduijn 2019). In general, the relevance of the rhetorical contestation of various contentious buzzwords in the parliamentary and the public arena, such as 'populist', 'realist' or 'terrorist', has been repeatedly underlined (Baele et al. 2019;Casiraghi and Testini 2021). As argued by both Finlayson (2007) and Krebs and Jackson (2007), rhetorical contestation is a central part of politics that concerns crucial controversies about meanings, ideologies and policies. Here, political actors find appealing incentives to contest controversial terms both to support their positions and to attack their competitors (Jamieson and Taussig 2017).
A few studies have analysed rhetorical attitudes towards populism. Elmgren (2018) has shown how the Finnish Rural Party has defended a non-mainstream and positive interpretation of populism, whereas Bale et al. (2011) addressed how the UK media has construed and employed the term. In addition, Casiraghi (2021) has investigated how British politicians have rhetorically referred to populism in the British Houses. Nevertheless, no study to our knowledge has so far focussed on the determinants of different rhetorical uses of populism in various countries.
Populism, indeed, may be construed as a thin ideology that counterpoises the 'pure' people versus the 'corrupt' elite (Mudde 2004), as a discursive approach that employs a peculiar anti-status quo rhetoric (Laclau 2005: 64-7;Panizza 2005: 2-3), or as a political style -a performative act to create a new category of 'people' (Moffitt and Tormey 2014). Then, political actors may present populism in a positive light, stressing their attention to people's true needs, or highlight some negative aspects, such as heinous characteristics of the far-right or simple demagogy (Caiani and Graziano 2019;Canovan 1981;Norris and Inglehart 2019;Ostiguy 2017;Stavrakakis 2017). Finally, the rhetorical description of populism may focus on different topics, from immigration to the EU and from technocratic politics to crime, may employ specific adjectives, such as cheap, nationalist or useful, and may target various actors, from far-left parties to the media or voters (Havlík 2019;Mudde 2007;Shehaj et al. 2021).
We will first present an overview of how politicians describe populism, highlighting the topics that emerge, the adjectives employed and the actors targeted. For instance, we expect that the debate in the UK is dominated by Brexit (Baldini et al. 2020;Ruzza and Pejovic 2019), the recent immigration crisis significantly influences the discussion in Italy (Albertazzi and Mueller 2013), and Catalan independence trumps most other debates in Spain (Barrio et al. 2018). Regarding the targets of the accusations, in Germany one might expect most attacks to be directed against leftist parties, especially Die Linke, though the recent success of the Alternative for Deutschland (AfD) might redirect such animosity towards the radical right (Arzheimer 2015). Conversely, populist parties like Movimento 5 Stelle (M5S) or Podemos might be expected to attract the majority of criticism in Italy and Spain respectively (Caiani et al. 2021;Mosca and Tronconi 2019;Sanders et al. 2017). Most importantly, our aim is to produce a comprehensive framework to investigate why politicians characterise populism negatively -e.g. 'rising right-wing, nationalist, and racist populism' (Congreso (2017b) or positively -e.g. 'populism means a dedicated attention to people's needs ' (Camera (2019b).
First, political ideology in its standard left-right form may have an impact on politicians' incentives to describe populism in different rhetorical terms. As argued by scholars who coined the expression 'populist hype' (Glynos and Mondon 2019), actors in the public and parliamentary arena in Europe tend to more often associate populism with right-wing, and especially far-right, parties. Moreover, such identification is consistently characterised in negative terms, where populism is linked with nationalist, racist attitudes (Mudde 2007). Hence, although a number of leftist parties in Europe have been labelled as populist, such as Die Linke in Germany or the Labour Party under the Corbyn leadership, we might expect to observe more negative rhetorical postures directed towards the right end of the political spectrum (Taggart and Pirro 2021). By contrast, as shown by Elmgren (2018) and Casiraghi (2021), right-wing politicians may adopt a negative perspective on populism, refusing and possibly redirecting accusations, or instead re-interpreting the concept, highlighting its positive values and proudly claiming the term for themselves.
On the other hand, another stream of the literature has argued that the impact of ideology on populism may not be linear. For instance, studies have found a U-shaped relationship between ideology and populism investigating both social media and party manifestos (Ernst et al. 2017;Rooduijn and Akkerman 2017). Since many populist parties have sponsored radical ideologies on both sides of the spectrum, an alternative hypothesis, which would run against the 'populist hype' argument, would be to expect a more negative rhetorical interpretation of populism at the centre. Conversely, at the extremes of the political field, one might expect parties to tend more often to support a non-negative characterisation of populism. Strategic motivations can also theoretically back this expectation: moderate, centrist parties should find more powerful incentives to contest far-left and far-right parties because of their radicalness, a feature that can be easily associated with populism and hence be stigmatised. H1a (linear): Since populism is more often negatively associated with right-wing parties, negative rhetorical attitudes increase moving towards the left end of the political spectrum.
H1b (U-shaped): Since populism is more often negatively associated with extreme parties, negative rhetorical attitudes increase moving towards the centre of the political spectrum.
Second, scholars have claimed that populism might constitute a socio-political reaction against specific phenomena (Kriesi 2014). Joppke (2021) has shown how the emergence of populism is often construed as a reaction against two 'liberalisms', namely multiculturalism and economic neoliberalism. One might then expect the positions of MPs to reflect such strife. Although Joppke (2021) argued that 'multiculturalism is central to populist mobilization', whereas the inequalities generated by economic neoliberalism and globalisation play a less important role, other scholars have underlined the connection between neoliberalism and populist reactions of revolt against it (Franzese 2019;Putzel 2020).
If these considerations are correct, we should observe more negativity towards populism from political parties that hold a positive view of multiculturalism and neoliberalism. By contrast, parties that oppose those policies should have stronger incentives to ride the populist wave, providing a more positive rhetorical interpretation (Akkerman et al. 2014). In the four Western European countries we consider, the struggle following the emergence on populism was centred more intensely on multiculturalism than on liberalism (Taggart and Pirro 2021). Hence, we expect parties' positions on multiculturalism to have a stronger effect on the likelihood to observe rhetorical negativity, compared to parties' preferences on neoliberalism.
In addition, populist parties across Europe have shown a significant degree of variation in their preferences in terms of economic policies. For instance, some populist parties like the AfD in Germany or ANO in Czech Republic have supported a fairly neo-liberal agenda whereas others, such as the M5S in Italy, have sponsored a more state-centred approach to economic affairs. Hence, predictions on the effect of the neoliberal variable should generate cautious expectations.
H2: Since populism is a reaction against multiculturalism, the more a party opposes multiculturalism, the more it positively characterises populism.
H3: Since populism is a reaction against neoliberalism, the more a party opposes neoliberalism, the more it positively characterise populism.
Next, the spread of anti-elite sentiments among the public played a major role in boosting the popularity of citizen-led grassroots movements (e.g. 'Occupy Wall Street' and 'Indignados'), as well as populist parties. A wealth of research has underlined how populist discourses focus primarily on the opposition between corrupt elites and morally upright people (Jagers and Walgrave 2007;Mudde 2004). In many countries, populist parties were indeed able to expand their electoral support, leveraging the need to wipe out the old, harmful elites from government, newsrooms or financial districts (Caiani and Graziano 2019;Hameleers et al. 2018).
More specifically, the moral corruption of elites is often the distinct target of populism. Populist political parties exploit (mis)behaviour in political affairs to highlight the supposed corruption of the governing elites (Curini 2020;Marx and Nguyen 2018). Consequently, parties that show a low propensity to employ an anti-elite rhetoric and denounce political corruption less often are expected to characterise populism more negatively, whereas the opposite should happen for parties that adopt anti-elite postures more frequently and that more regularly highlight the importance of reducing political corruption. 2 H4: Since populism is often characterised by an anti-elite rhetoric, the more salience a party puts on anti-elite discourses and on the necessity to end political corruption, the more it will positively characterise populism.
The expectations we have spelled out may vary depending on the topic that politicians are discussing when they mention populism. For instance, during debates about immigration, right-wing parties should be more prone to praise populist policies or slogans supporting their tougher stance on integration. On the other hand, when discussing economic matters, leftist parties should be more incentivized to exploit a positive outlook on populist positions that include expansive and supposedly fiscally irresponsible policies. Finally, when discussing more general topics, such as the problems and characteristics of the political system, ideology will unlikely affect the outlook on populism. Since our aim is to investigate carefully the detailed rhetorical manifestations of the parliamentary debate on populism, we will test our models considering separately the most common topics in our corpus.

Data and method
We focussed on four Western European countries to exploit variation on different levels. In Germany, accusations of populism reached both left (Linke) and right-wing parties (AfD), and the debate on populism has focussed on both economic and immigration issues (Arzheimer 2015). In Italy, two different populist parties (M5S and Lega) have enjoyed broad popular support, participating in Italian governments with major roles, and putting emphasis on often divergent policy issues (Albertazzi and Mueller 2013). In Spain, Podemos has adopted an institutionalised role, following a different strategy compared to populist homologues in other states. Likewise, a radical-right populist party, Vox, has recently enjoyed significant electoral success, and the path towards independence of regionalist entities has been marked by discussions centred on populism (Barrio et al. 2018;Sanders et al. 2017). In the UK, no members of (explicit) populist parties (except for two UKIP politicians) have held seats in the Houses. Still, populism-related issues like Brexit have been debated at large, and the Conservative Party under the leadership of Boris Johnson has received a number of accusations of populism (Casiraghi 2021). Investigating such diverse political environments, our analysis will shed light on different developments of the debate on populism.
We focus on the 2010-2020 period for one main reason. While the meaning and interpretation of populism has repeatedly changed in the eyes of academics and the public (De Cleen and Stavrakakis 2017), we would expect that over a relatively short period of time politicians should hold at least a basic shared understanding of what populism is. Tackling politicians' rhetorical divergences on populism, we collected all paragraphs (N = 4.835) in which MPs mention the keywords 'populism' and/or 'populist', manually searching through the Parliamentary Archives of the four countries. We do acknowledge the limitations of our selection, as individual and party-determined biases might drive some politicians to engage explicitly with the terms, and others to refer to them implicitly. At the same time, there are no reasons to assume that such potential distortions would not occur if we were to include implicit mentions as well. In particular, as there are linguistic differences in how to handle implicit references, the opposite strategy might introduce biases at the country-level. Moreover, our main interest was to investigate how politicians 'talk' politics, further reinforcing the preference for explicit mentions.
Then, we manually coded all paragraphs to assess the topics that characterise the debate, the adjectives that politicians use to describe the keywords, and the political actors whom they target. Although studies on political rhetoric have focussed on detailed rhetorical analyses, such as the distinction between different rhetorical strategies and arguments (e.g. Casiraghi and Testini 2021; Walter 2017), we will consider the most basic yet fundamental element of the debate on populism: the negative versus non-negative interpretation of the term.
In order to assess such rhetorical contestation, we relied on manual coding (Krippendorff and Bock 2009). In particular, we carefully read every paragraph, assessing whether the reference to populism was positive, neutral or negative. 3 Then, we merged the neutral and positive observations together to obtain a dichotomous variable. Considering the pronounced negative stigma associated with populism, both neutral and positive references clearly constitute a powerful signal that an MP provides about their non-mainstream, non-negative interpretation of populism. 4 Indeed, we expect MPs to refer to populism in neutral terms strategically, to avoid the associated stigma, even when they actually imply that populism is inherently a 'good thing' .
Our decision to employ a manual approach to text analysis derives from established practices in political Science literature. First, we wanted to produce a detailed picture of the debate on populism, uncovering the rhetorical dynamics that define the parliamentary discussion, a goal that required a qualitative approach (Finlayson 2007). Second, regarding the assessment of negative references, we have also decided to resort to manual coding. To ensure that manual coding represents the best option for our purposes, we run a dictionary-based model, taking into consideration the multi-lingual nature of our corpus: results confirm that the automated analysis underperforms compared to human coders. 5 We focussed on party-level variables to test our hypotheses, as parties are the engine of parliamentary activities and strongly shape the positions and rhetorical strategies of their members. In particular, we employed various waves (2010, 2014, the 2017 special wave and 2019) of the Chapel Hill Expert Survey to assess parties' positioning on left-right ideology, multiculturalism, economic interventionism and protectionism, anti-elite rhetoric and corruption. 6 Finally, we collected data that serve as control variables. First, attitudes towards populism may be explained by strategic incentives. Since populism is considered to offer naïve solutions to complex issues, government members might be incentivized to dismiss alternative policies proposed by the opposition as populist. Second, as populism is generally construed as a hot and negative issue, an upcoming election may shape politicians' incentives to negatively describe populism, exploiting it as a rhetorical expedient to criticise competitors during the electoral campaign. Similarly, if the characterisation of populism has become more unfavourable as time has gone by, recent observations may show more negativity. In addition, since an unfavourable position towards European integration is often associated with populist stances, pro-EU MPs might be more likely to characterise populism as negative compared to their Eurosceptic colleagues (Rooduijn and van Kessel 2019). Finally, we control for how much a party is considered populist, assuming that populist parties might be more incentivised than non-populist parties to re-construe populism as a positive or at least neutral concept to avoid being stigmatised (Norris 2020). Table 1 offers an overview of all variables.

Results and discussion
Before discussing our statistical results, we present a description of the rhetorical characteristics of the debate on populism in the four countries. Table 2 shows the most common topics, the most frequent adjectives and the most recurrent political actors that MPs target.
The topics that politicians discuss more often are common political matters, such as political equilibria in the assembly, economic issues like financial reforms, and EU politics. In Germany, many debates on populism focussed on immigration, especially after Merkel's 'Wir schaffen das' speech about the willingness of the Government to accept a large number of asylum applicants (Oltermann 2020). Opposition to this 'welcoming culture', especially from the AfD, led the Government to counteract such criticism by highlighting its supposed populist character. The most used adjectives, such as 'right-wing', 'racist' and 'nationalist' appear to reflect this parliamentary and public strife.
In Italy, crime, political corruption and justicialism were popular topics, the first two often mentioned by the M5S and the latter brought up by the right. Relatedly, one of the most frequently used adjectives is 'penal', along with other more common adjectives, such as 'dangerous' and 'demagogic' . In this context, right-wing parties systematically accused the left and the judicial system of populism, following the latter's supposedly populist campaigns against leaders such as Berlusconi (Dallara 2015). On the other hand, mainstream parties reacted with accusations of populism against the M5S' supposed obsession with political corruption, often referring to its populist anti-political stances. Along with references to the populism of Italian society, the most frequent targets are the M5S, a party considered populist since its foundation, and the Lega, more recently identified as populist. In Spain, debates mainly focus on Catalan independence, and broadly on the regional political turmoil against the central Government. Local politics was a significantly frequent topic, the adjective 'nationalist' figured prominently and independent parties and regional authorities were among the most common targets. In addition, both Podemos and the Partido Popular (PP) often faced the accusation of populism, mostly for being supposedly demagogic about internal political matters and the relation with the EU, whereas Vox did not figure among the most frequent targets as it entered the Spanish Assembly only recently (Turnbull-Dugarte 2019).
In the UK, Brexit and EU politics characterise many debates centred on populism. The most common adjectives, such as 'nationalist', 'dangerous', 'right-wing' were consistently employed to attack the Conservative Party, UKIP and voters supporting Brexit. Conservatives reacted to such accusations by employing populism themselves as a rhetorical weapon, highlighting the demagogic and populist approach to economic reforms of the Labour Party. Although politicians employed populism as a rhetorical expedient also while debating other issues, such as the Conservatives' tough take on crime or the Labour Party's soft approach on immigration and on Northern Ireland, Brexit dominated the discussion (Baldini et al. 2020).
Next, in order to assess our formal hypotheses, we fit a series of Mixed-Effects Binomial Logistic Regression models, investigating the determinants influencing the probability of speaking of populism in positive rather than in negative terms. We opt for a Mixed-Effects specification to take into consideration the hierarchical nature of our dataset -observations nested within parties nested within countries -specifying a random intercept at the country level and a random coefficient effect for parties. In the Online Appendix, we provide a more detailed discussion of the methodology and present several robustness checks, finding that our results are robust for different specifications. Table 3 presents our models: Models 1 and 2 test our hypotheses on all observations, while Models 3 to 6 assess whether the topic under discussion influences how different dimensions shape the way MPs characterise populism.
As shown in Model 1 and Model 2, harsher positions on multiculturalism significantly increase politicians' likelihood to produce positive rhetorical references towards populism, and a similar effect characterises the emphasis on anti-elite rhetoric and corruption. These findings support H2 and H4: politicians who favour multiculturalism and do not resort to anti-elite or anti-corruption rhetoric more consistently criticise the populist character of such positions. For example, in the UK, Conservatives were often under attack for their supposedly hypocritical anti-elitism: 'the populist right […] speaking for the people against the elite. Noble Lords need only look at Jacob Rees-Mogg or Boris Johnson himself to see how ludicrous their claim to represent the people against the elite is' (House of Lords 2019). Regarding multiculturalism, MPs who favour more multicultural societies can easily accuse their adversaries of populism, linking positions against inclusion to typical populist, nationalist postures. For example, in Spain, populism was explicitly connected to xenophobia and even Nazism, as it 'drives what we believed we had buried in Nuremberg' (Congreso 2017c). On the other hand, even if in general the more a party opposes multiculturalism the more its MPs will consider populism in positive terms, many debates show the rhetorical flexibility of populism as a conceptual weapon to contest competitors' policy preferences.
Indeed, MPs can as easily argue that the supposedly soft approach to immigration adopted by their opposition is naïve and populist, as it does not consider the actual socio-economic situation and citizens' preferences. For instance, left-wing Spanish parties often faced the imputation of exploiting 'populistically' the deaths of immigrants and refugees to reject tougher stances on immigration (Congreso 2018). In particular, the PP warned against the populist and demagogic requests for 'documents for everybody' and highlighted the importance of 'secure borders in Europe' (Congreso (2017a).
Next, to test hypotheses H1a and H1b we explore the effect of parties' positions on the left-right scale on the characterisation of populism. H1a is not supported, as we do not find any significant linear relation between political ideology and the characterisation of populism. Such a result contradicts previous studies that highlighted the tendency to massively characterise right-wing and far-right parties as populist, the so-called 'populist hype' (e.g. Traverso, 2019;Vieten and Poynting 2016). Evidently, politicians 'see' a lot of populism also on the left of the spectrum, and ideology does not seem specifically to incentivise MPs to characterise populism in negative or positive terms. Such a result also suggests that, at least explicitly at a rhetorical level, right-wing parties have not been much affected by a supposed populist contagion, as they systematically persist in a negative interpretation of populism. In Italy, notably, accusations of populism are directed at both right and left-wing parties: the Lega was attacked as the promoter of 'right-wing populism' that on immigration ruthlessly 'exploits a fear that is more perceived than real' (Camera 2019a). On the other hand, however, M5S's leaders were referred to as 'professionals of anti-politics, without a serious party to speak of, built on populist slogans' (Camera 2010).
In Model 2, we investigate the quadratic relation between the left-right scale and the rhetorical contestation of populism. As suggested, parties with more extreme positions might consider populism with a more positive attitude, compared to centrist ones that are less often accused of populism and, thus, might use 'populism' to stigmatise their opposition. Even if we do not find strong empirical support for H1b, in Figure 1, we plot the expected values of the dependent variable for each value of the left-right spectrum. We find evidence that parties at the extremes of the spectrum are, on average, slightly more likely to characterise populism in positive terms, compared to their centrist homologues. 7 Such results are of particular interest, as previous research suggested both that populism should be prevalent among right-wing parties and that extreme parties are more likely to adopt a populist rhetoric in their social media posts (Ernst et al. 2017). Hence, our findings suggest a marked difference between how politicians 'talk politics' and how they talk 'of ' politics. Finally, results show non-significant results for economic variables, hence not supporting H3. As claimed, such a null result may be due to the considerable variation in terms of economic policy preferences of populist parties across Europe. In addition, some control variables show a significant effect: first, the more populist a party the more it tends to interpret populism in positive terms. As there is still an open debate about the rhetorical strategies of populist politicians (Casiraghi 2021;Elmgren 2018) -whether they accept or refute the definition of populist-this result proves that populist parties tend to construe populism more positively. Second, rhetorical negativity towards populism increases as time goes by, possibly suggesting a more negative interpretation of populism in recent times, especially after highly mediatised populist phenomena like Brexit or the European power struggle on immigration.
Furthermore, considering populism's fluid nature, the topics of parliamentary discussion might influence the form of the rhetorical contestation. Thus, in order to explore a largely uncharted territory and to investigate thoroughly the effect of ideology on the characterisation of populism, we fit Model 2, considering debates on the four most common topics: economics, the political system, the EU and immigration. 8 Across the four Models, results present some interesting divergences. Model 3 (economics) shows that even when the topic is economic policies, H3 is not supported. Besides the great variation in terms of economic preferences of populist parties, a look into the debates further explain this lack of significant correlation. For instance, in Germany, where debates on populism and economic affairs are more common compared to other countries, both domestic and foreign actors advocating the acceptance of more debt were harshly criticised and accused of populism. In 2018, the Minister of Finance faced backlash as he was not 'ensuring that the European Commission initiates an excessive deficit procedure against the populist Government in Rome' (Bundestag 2018). Furthermore, parties like the AfD, which holds a neo-liberal stance regarding state interventionism, adopted at the same time a non-negative attitude towards populism. It would seem that the AfD either referred to populism to dismiss the accusation of populism as ridiculous: ' A ghost is haunting Europe: populism. He can be found everywhere. However, nobody knows who he actually is' (Bundestag 2018), or to adopt the opposite strategy, claiming the term as something to be proud of: ' Actually choose populist; because we represent your opinion' (Bundestag 2019).
Next, Model 4 (the political system) shows that when debates address internal political affairs, the most important drivers that explain the direction of the rhetorical contestation are the populist character of the party, anti-elite stances and government membership. This is not a surprising result, given that supposedly populist, anti-elite positions have dominated the internal politics debate since the strong emergence of populist parties in most European countries. One additional result emerges from Models 5 (EU) and 6 (immigration): harsher stances on multiculturalism imply a more positive characterisation across all models (except Model 4), testifying the importance of the link between multiculturalism and populism, regardless of the topic under discussion. In particular, multiculturalism appears to be particularly relevant in shaping the debates on the European Union, to the point that even when the debate concerns the EU, multiculturalism is more relevant to the interpretation of populism than parties' positions on the EU.
Besides these considerations, the most intriguing aspect of the comparison across the four Models is the relation between political ideology and the rhetorical contestation of populism. Model 3 shows that, when debating economics, moving towards the right of the spectrum the expected value of the dependent variable approximates zero. Conversely, parties at the left extreme tend on average to consider populism in more favourably, as shown in Figure 2. For example, a Die Linke MP asked their colleagues to 'explain what is populist about it when you find out that people request that the law ensures equal wages for the same work' (Bundestag 2016b). Likewise, an M5S deputy suggested that 'when a retired man is earning 485 euros per month, that is not populism, that is the real world' (Camera 2015).
A similar trend characterises the debate on the EU. Figure 2 shows a slightly U-shaped relation: when the debate focuses on the role of the EU, extreme parties tend to support a positive interpretation of populism, likely reflecting the anti-EU stances of their voters. Differently, centrist parties accuse their 'extreme' colleagues of populism, offering a consistently negative interpretation of such attitudes in relation to the EU. Conversely, Figure 2 also suggests that leftist parties seem to characterise populism more negatively compared to their colleagues from the right when the topic is immigration.
The contrast between the impact of ideology when the debate concerns economic affairs and immigration can explain the slightly curvilinear effect observed in Model 2 and Figure 1. Extreme-left parties characterise populism negatively and employ it as a rhetorical weapon to contest the anti-immigration stances of their far-right colleagues, who, on this debate, tend more often to highlight the positive characteristics of populism. As said, right-wing parties usually argue that the debate on immigration should be less 'idealistic' and that politicians should listen more to the people's needs on these matters. By contrast, right-wing MPs employ more consistently a negative interpretation of populism to warn against the economic preferences of leftist parties who favour a stronger state control on the national economy and more strenuously oppose neoliberal positions.
Such results offer interesting implications for the academic debate on populism and in particular on the 'populist hype' . Following the recent, heated public debate on immigration, a discussion where populism was often called into question, media and scholars alike have repeatedly highlighted the connection between populism, nationalism and anti-immigration stances. Evidently, however, the manifestation of populism and its rhetorical use in the parliamentary arena have followed a different trajectory. Most often, accusations of populism are directed towards the left, and the same leftist parties seem to have a more positive view of populism, at least in the countries under scrutiny. Hence, scholars and the public should be cautious in systematically linking populism and far-right stances, especially when assessing how politicians interpret and rhetorically charge the terms 'populism' and 'populist' .

Conclusion
We have addressed an overlooked perspective of the otherwise vast literature on populism. Political Science has largely discussed populism's characteristics and effects -to the point that the concept is considered the 'buzzword' of recent years. However, there is scant research on how politicians talk about populism, a crucial gap our article has tried to fill. Among the relevant topics, we find corruption and justicialism in Italy, and local politics in Spain, linked to the regional turmoil against the central Government. Immigration stands out in Germany, whereas Brexit and the EU are the focus of many debates in the UK.
Our results provide support for two of our hypotheses. As both anti-elitism and the salience of corruption are core characteristics of populism, we find that a lower score on these scales determines a more negative view. More interestingly, harsher positions on multicultural policies increase the likelihood of having a positive attitude towards populism, across all topics. The link between populism and immigration is so crucial, both for the public and the parliamentary discussion, that multiculturalism emerges as a relevant driver of the rhetorical interpretation of populism, also when the topic of the discussion is mainly the EU or economic affairs.
In addition, our article has shed light on two (contrasting) hypotheses, the so-called 'populist hype' and the expectation that extreme parties would more likely characterise populism as positive. We show how parties talk about populism differently depending on the topic, with right-wing parties more likely to present populism negatively when debating economic affairs and the EU. Thus, parties understand and consider populism in a way that differs significantly from how academics and the media portray it.
Our article further contributes to the literature in a number of ways. First, it expands our understanding of populism through the lens of politicians rather than political scientists. Second, it contributes to the literature on rhetorical strategies, assessing the role of ideology on the use of relevant buzzwords, and especially how its effects vary depending on the topic. Next, it offers insights into how parties position themselves on the debate on populism. Finally, we have constructed a rich, manually coded dataset that can be exploited for future research.
To conclude, we suggest future lines of research that are beyond the scope of this article. First, the analysis could extend geographically, including countries with more marked differences in terms of political regime and history. Likewise, future studies could focus on a more extended period of time, assessing the evolution of the meaning of populism across different periods; a research design that could be appropriate for innovative text analysis techniques, such as word embedding (e.g. Rodman 2020). Finally, a related stream of research could highlight more specifically the 'buzzword' dimension of populism, by proposing a comparison with other domains, such as social media.

Notes
1. An additional 'take-away' of our dataset is the validation of using Google Translator for research (see Supplementary materials). 2. Anti-elite rhetoric and corruption can theoretically pose an issue of endogeneity: a party with a strong anti-elite stance is likely to be populist and may consequently frame populism more positively. However, there is still discussion in the literature about whether populist parties do consistently adopt a positive interpretation of populism (e.g. Casiraghi 2021; Elmgren 2018). Moreover, we statistically check for the potential endogenous effects of both anti-elite rhetoric and corruption and we do not find evidence of endogeneity (Online Appendix). 3. In Online Appendix, we provide various examples of coded paragraphs and a complete codebook. 4. Results remain robust when we employ the dependent variable with three categories. 5. For further details on possible alternative methodologies, see Online Appendix. 6. https://www.chesdata.eu/our-surveys 7. For further discussion see Online Appendix. 8. We selected the most mentioned topics considering all countries.
Furthermore, for space reasons, we do not present models testing H1a for the topics, as results are unchanged with respect to the general model (Model 1).
Pagano, Francesco Piacentini and Valerio Vignoli for their support and their valuable comments and feedback on our manuscript.

Disclosure statement
No potential conflict of interest was reported by the author(s).