Sage Journals: Discover world-class research

Abstract

Insurgent candidates from across the political spectrum are increasingly turning to social media to directly engage the public. Social media offer a platform that favours affect and personality, both key components of populist-style rhetoric, a label that is often attached to politicians outside the political establishment. Despite noteworthy exceptions, few cross-national studies of high-profile candidates’ use of social media exist, and even less is known about how candidates representing various political ideologies employ affect alongside populism. To advance the state-of-the-art, this study examines the sentiment and rhetorical targets of attack in the Twitter feeds (N = 25,825 tweets) of six presidential candidates in the United States and French election campaigns of 2016 and 2017. Employing dictionary-based quantitative analysis, the study finds variation among the candidates’ rhetoric in terms of how they employ populist themes, affect and ideology. The findings suggest that scholars should consider a more nuanced approach to populism in late-modern democracies.

Keywords

Automated content analysis comparative France populism populist communication presidential elections Twitter US

Recent scholarship has examined the use of populist-style rhetoric in party leaders’ social media feeds in both European and American democracies (e.g. Ernst et al., 2017; Zulianello et al., 2018). So-called ‘populist’ candidates have successfully leveraged these platforms to elevate discourses of anti-elitism to various degrees. Accordingly, scholars now frame populism as an issue of political communication, ushering in a so-called ‘discursive turn’ in populism studies (De Vreese et al., 2018). Given the relevance of social media to building political movements analysing the discourse¹ of major candidates on platforms like Twitter and Facebook may provide a more detailed assessment of the ideological valence of populist movements (Bonikowski and Gidron, 2016; Engesser et al., 2017; Ernst et al., 2017). Yet, with notable exceptions (Ernst et al. 2019), few studies have empirically confirmed the various components of populist-style rhetoric on social media . Building on previous comparative studies (Ernst et al., 2017, 2019; Zulianello et al., 2018), this article investigates the communication of new, highly relevant cases of populist movements and the political figures representing them on social media. Results demonstrate that scholars ought to distinguish different subtypes of populist communication.

Summarizing the first generation of empirical studies of social media-based populist discourse, Hameleers (2018) points out that ‘this one size fits all approach to populism may not be accurate, considering the large differences among and within continents’ (p. 2184). In response, this study asks whether we can empirically distinguish between ideological subtypes of populist discourse on social media, given that politicians who employ it have different political convictions. It has long been argued that populism is an empty shell that can be filled with different ideological components (Mény and Surel, 2001), but surprisingly this has rarely been empirically demonstrated. Thus, we see a need for more comparative studies that explicitly explore the ideological nuances of populist discourses within and across cultural bounds. Focusing merely on the distinction between populist and non-populist communication ignores the fact that populism is often combined with substantial ideological convictions, and, therefore, represents a heterogeneous set of values, motivations and potential outcomes.

To address this research gap, we analyse the Twitter feeds of six candidates who ran for office in the French and United States presidential elections: Donald J. Trump, Hillary Clinton, Bernie Sanders, Emmanuel Macron, Marine Le Pen and Jean-Luc Mélenchon. They represent different ideological streams in modern politics and have an exceptional visibility beyond national boundaries. Data for this study come from a comprehensive collection of candidates’ official tweets during a 13-month window around respective election days (8 November 2016 and 7 May 2017; N = 25,825 tweets). Drawing on qualitative and quantitative discourse-centred studies in this area (Bonikowski and Gidron, 2016; Jagers and Walgrave, 2007; Oliver and Rahn, 2016; Rooduijn and Pauwels, 2011; Stockemer and Barisione, 2017), the analysis is based on custom dictionaries built for capturing elements of populist rhetoric in the form of verbal references to known political targets: ‘the elite’, ‘the people’, and perceived threats to the in-group (immigration rhetoric). Furthermore, sentiment, a key component in social media appeals, was measured to capture tone (Young and Soroka, 2012) relative to the targets of populist discourse in the posts.

Based on recent social and economic developments, as well as structural similarities between the election systems in France and the United States – both countries have direct voting for candidates, stark political polarization, the emergence of candidates that can be classified as insurgents (Gerstlé and Nai, 2019; Hawkins and Rovira Kaltwasser, 2017), and the adoption of Twitter by campaigns – the respective election campaigns are good cases for comparison in a cross-cultural study. At the same time, the divergences in terms of political culture that exists between the contexts provide a robustness check for our assumption that populist Twitter discourse surfaces in ideologically coloured subtypes across countries and cultures.

Elements of populist discourse

Populism is a ‘thin’ political ideology that manifests in discourses that elevate conflict between the perceived ruling elite on the one hand, and the will of the people on the other hand (De Vreese et al., 2018; Mudde, 2004). The discourse revolves around the categories of ‘the elite’ and ‘the people’, which are central to how populism understands democratic politics (Canovan, 1999; Mény and Surel, 2001; Mudde, 2004). Both are flexible concepts. In a populist vision, the notion ‘elite’ includes all dominant institutions and decision makers in society (Laclau, 1979), be that in the domain of politics, the economy or cultural production. The notion can stretch to include the entire, allegedly dysfunctional political order, often pejoratively termed ‘the system’. According to Canovan (1999), there are several notions of ‘the people’ as well. For example, an ethnicity-based definition, which is employed in nationalist right-wing populisms, tends to see immigrants as social outsiders. Mény and Surel (2001) claim that many populist parties have fallen into this mould, making immigration their central issue.

Simply referring to ‘the people’ without reference to elites represents populist discourse in its most basic form (Jagers and Walgrave, 2007: 323). In contrast, the joint occurrence of references to the people and either attacks on elites, or a critique of immigration, yields richer rhetorical forms termed ‘anti-elitist’ and ‘excluding’ populism. Finally, the combination of all three elements in a political actor’s discourse is called ‘complete populism’ (Jagers and Walgrave, 2007: 335). Developing this typology further, Hameleers (2018) points out that populist narratives vary, especially in regard to the anti-elitist dimension: They can target different groups of elites, such as political, economic, or cultural. We assume that these variants in anti-elitist discourse correspond with the ideological background of the communicator, leading to not one, but several forms of populist narrative.

The discursive turn in populism research means that it is not necessary to determine beforehand which politician is a populist-based on party platforms or campaign documents. Rather, it can be ascertained empirically based on their communication. Yet in practice, as Stanyer et al. (2016: 361) point out, ‘our knowledge is poor about how frequently both populist and non-populist mainstream political actors refer to the people, express anti-elitism, and exclude various outgroups in their communication’. However, given that political communicators can be more or less clearly populist at different times, the distinction between populist and non-populist actors is less meaningful than the distinction between politicians belonging to opposed ideologies. While political outsiders who challenge the establishment might be more inclined to use anti-elitist rhetoric in general, ideological orientation might account for who the predominant targets of attacks on ‘the elite’ are. The rhetorical choice of targets thus reflects the proposed social and political values behind the particular variant of populist discourse.

In addition to the binary rhetoric of power holders versus ordinary people, affect is also a standard feature of populist communication. Affective rhetoric depicts the polity in a state of crisis, and uses emotionalized blame attribution to attack the elite and/or immigrants (Hameleers et al., 2017; Taggart, 2004). Appeals to fear and anxiety undermine rational debate and re-enforce tribal alliances through identity cueing (Canovan, 1999: 6, Diehl et al., 2019; Wodak, 2015), yet scholarship has mostly ignored sentiment from an empirical perspective. Thus, ‘elite’ versus ‘the people’ appeals should coincide with both negative tone (in reference to elites) and positive tone (in reference to the people).

Populist discourse on social media

Social media provide an ideal communication space to express the aforementioned ideas to target groups (KhosraviNik, 2018). Twitter, in particular, allows political communicators to bypass journalistic filters and easily connect with the public through attacks on the defined opposition. Therefore, it is not surprising that an ‘elective affinity’ of populism and social media has been diagnosed (Gerbaudo, 2018: 745). Due to the growing importance of social media for political discourse directed towards the general public, it is not surprising that researchers are increasingly scrutinizing the social media practices of politicians (Bracciale and Martella, 2017; Ernst et al., 2017; Zulianello et al., 2018). The architecture of these platforms favours the politics of affect and personality over deliberation, because they elevate visibility and affirmation.

For example, one of the first transnational, qualitative studies of populism on social media revealed that elements of populism are present in both the tweets and Facebook posts of a diverse set of European politicians in terms of prominence and party affiliation (Engesser et al., 2017). Engesser at al. (2017) suggest that populist rhetoric occurs in a ‘fragmented form’ (p. 13); the messages contain only partial or very specific elements of the full discourse (e.g. references to the people may appear alongside or completely detached from anti-elite references). Ernst et al. (2017) focused on the Facebook and Twitter accounts of 88 politicians from six Western democracies, finding that politicians from extremist parties include more populist ideas in their messages, and the same holds true for politicians of opposition parties.

Populism occurs in the tweets of politicians from right- and left-wing parties, although it is more often combined with elements of right-wing than left-wing discourse. In their analysis of Facebook postings from 2015, involving democracies from Europe, North and Latin America, Zulianello et al. (2018) found that only a handful of European right-leaning politicians, including Marine Le Pen, Farage and Berlusconi, are examples of complete populism. A larger group of politicians who lean left on the political spectrum employ anti-elitist populism (but devoid of ethno-centric language), while most politicians used no populist discourse at all.

Subtypes of populism

Previous scholarship established the scope of populism on social media platforms and the distribution between different types of politicians. This study argues that populist rhetoric can be further differentiated in terms of the targets of attack and affect (or sentiment). Based on the idea that populism is a flexible vessel that can be combined with a host ideology (Gidron and Bonikowski, 2013; Hawkins and Rovira Kaltwasser, 2017; Mény and Surel, 2001), we expect that populism supplements existing ideological tendencies and dresses them in a specific rhetoric. Thus, the political background of the selected politicians is expected to matter for their application of populist framing in social media campaigns .

Overall, the frequency of references to ‘the people’ (i.e. people-centrism) and ‘the elite’ (anti-elitism) indicates varying tendencies towards populist rhetoric in general. However, we do not expect uniformity concerning the choice of targets across the six candidates analysed in this study, given their different ideological provenience. Concretely, we expect candidates to vary with regard to the tone and targets in their social media messaging, and this variation should cohere to their ideological tendencies.

Candidates who are closer to a left-wing, socialist ideology, which sees economic inequality as the main political problem facing society, should target the economic elite more often than other candidates. However, candidates closer to a right-wing, conservative-nationalist ideology, which advocates pushing back against neo-liberal and post-materialist values in the political system, should target the political and cultural elite (e.g. the media elite and/or sitting politicians) more often. Right-wing populism defines the elite differently than left-wing candidates, and they should attack political institutions instead of the economic order.

Since conservative and nationalist politicians perceive immigration to Western countries as problematic for reasons of national identity and cultural homogeneity, right-leaning candidates should offer the most complete populism. Furthermore, we expect that those candidates who represent a more radical alternative to the political establishment to cultivate a more people-centric rhetoric than moderate candidates closer to the political establishment, because speaking on behalf of the people is a means of legitimizing an outsider’s claim to power. These arguments can be translated to concrete hypotheses:

H1: Candidates with a conservative or right-wing orientation discursively construct an opposition between the people and the ruling liberal elite, and thus target political and cultural elites and institutions more often than economic elites.

H2: Candidates with a left-wing orientation construct an opposition between the people and the economic elite, and thus target the economic elite and institutions more often than other types of elites in their tweets.

H3: Candidates with a right-wing orientation construct an opposition between the people and ethnic outgroups, and thus make immigration and the protection of borders a salient issue in their tweets.

H4: The discourse of insurgent candidates from the far-ends of the political spectrum is more people-centric than that of insurgents from the centre such as Macron.

Finally, the tendency of populists to ‘denigrate’ the elite and to ‘venerate’ the people (Stanley, 2008: 102) in their discourse should be stronger in combination with a more extreme ideological position:

H5: Candidates challenging the establishment from the left or right (Trump, Sanders, Le Pen, Mélenchon) should use a more negative tone when talking about the elite and a more positive tone when talking about the people than Macron who challenges the establishment from the centre.

Insurgent candidates in the French and United States presidential elections

The 2016–2017 American and French presidential elections provide the framework for the study of candidates’ social media rhetoric. Both campaigns were marked by the rampant rise of insurgent candidates across the political spectrum who presented themselves as leaders of new political movements, including Donald Trump, Bernie Sanders, Marine Le Pen, Jean-Luc Mélenchon and Emmanuel Macron. While these candidates can be seen as challengers to their party establishment or established parties in general, their ideological beliefs are markedly different. Among the American politicians, Donald Trump belongs to the conservative right (Oliver and Rahn, 2016), Bernie Sanders is a self-characterized left-leaning ‘proud socialist’ (Kazin, 2016), while Hillary Clinton belongs to the liberal-centrist camp. Trump and Sanders (as well as all French candidates) referred to their support base as a movement rather than a party underlining their outsider image.

The French politicians mirror this pattern: Marine Le Pen was the leader of the insurgent far-right movement Front national (Rassemblement national), Jean-Luc Mélenchon was the leader of the left-wing movement La France insoumise, and Emmanuel Macron cast himself as a liberal reformer from the centre with his movement En Marche!. Trump and Sanders have been called populists and outsiders, while these labels can also be attached to Le Pen and Mélenchon due to the location of their movements at the far-ends of the political spectrum in France, which has earned them a pariah status among the ruling establishment (Kazin, 2016; Reynié, 2011). However, Clinton and Macron are widely seen as moderate candidates (Escalona, 2017).

Twitter has come to play a crucial role in political communication in both countries (McGregor et al., 2017; Mercier, 2016; Wells et al., 2016), and political polarization has become stronger in recent years with new movements developing on both the left and the right. However, political cultures are unique; including cases from both contexts provides an important contextual variation that may strengthen the cross-cultural validity of subtypes of populism.

Methodology

Sample

Data for the study is based on a sample of tweets (N = 25,825 tweets) obtained from Twitter’s public application programming interface (API) search function. Several studies have employed this method (Borra and Rieder, 2014; Bruns, 2012; Vergeer and Hermans, 2013). Since this study is concerned with campaign rhetoric – and to avoid complications from overly large data sets – the period was narrowed to a 13-month window around the most recent presidential elections in France (7 May 2016–7 June 2017) and the United States (8 November 2015–8 December 2016). Twitter handles representing the official accounts of Marine Le Pen (@MLP_officiel), Emmanuel Macron (@EmmanuelMacron), Jean-Luc Mélenchon (@JLMelenchon), Hillary Clinton (@HillaryClinton), Bernie Sanders (@SenSanders) and Donald Trump (@realDonaldTrump) were used for collecting data. The data include the number of likes, re-tweets, replies and the complete text for each tweet (Table 1). Data were collected in July 2017.

Table 1.

Frequencies and percentages of all tweets by dictionary category and candidate.

	Political elite	Corporate elite	Media elite	Immigration	The people	Total
	n (%)	n (%)	n (%)	n (%)	n (%)	n
US candidates
Clinton	363 (7.4)	1160 (23.7)	49 (1.0)	96 (2.0)	357 (7.3)	4888
Sanders	429 (16.5)	739 (28.5)	34 (1.3)	47(1.8)	477 (18.4)	2593
Trump	1331 (27.5)	113 (2.3)	882 (18.2)	131 (2.7)	348 (7.2)	4841
French candidates
Le Pen	544 (14.7)	171 (4.6)	46 (1.2)	381 (10.3)	448 (12.1)	3703
Macron	202 (6.4)	103 (3.3)	6 (0.2)	43 (1.4)	140 (4.4)	3153
Mélenchon	634 (9.5)	225 (3.4)	63 (0.9)	86 (1.3)	260 (3.9)	6647

Note. N = 25,825. Tweets are counted with one or more dictionary category mention.

Instrument and method

To measure populism in the text of the tweets, we analytically separate the concept into references to the people, references to the elite and tone (sentiment). A higher number of references to either or both of the constructs ‘people’ and ‘elite’ indicates a populist tendency. Employing sentiment alongside populist terms, depending on the construct, is an additional indicator of anti-elitism or advocacy for the people. The analysis thus combines two dictionary-based approaches to automated content analysis: custom word lists to capture references to the people and different elite groups, and sentiment scoring (Grimmer and Stewart, 2013).²

First, in order to tap variations in how often the candidates refer to the constructs people and elite, two custom dictionaries (one in French and one in English) were created. The authors employed a systematic process of qualitative pre-analysis, comparison with existing word lists (Bonikowski and Gidron, 2016; Jagers and Walgrave, 2007; Oliver and Rahn, 2016; Rooduijn and Pauwels, 2011), and subsequent reliability testing of the dictionaries through several rounds of hand and computer-aided coding (Lewis et al., 2013; Su et al., 2017). The qualitative pre-analysis identified expressions that relate to populist cueing in the form of anti-elitism (including references to ‘the system’ and its emblematic institutions, actors and processes), people-centrism and the topical mentions of immigration in the tweets of each candidate. The qualitative step was necessary to identify the meaning behind specific terms in the vernacular employed by the candidates, yielding a variety of idiomatic terms.

Although one might argue that only pejorative and generalizing references to the elite, such as for example the word ‘oligarchy’, are strong indicators for populist anti-elitism, our measurement strategy separated between the salience of the constructs in the discourse and the tonality with which they were discussed. This reflects, however, that the salience of ‘elite’ terms in general is a key component of populist discourse (even when pejorative terms are not used ) and that ‘neutral’ words (broad references to political actors or institutions) may automatically acquire a negative connotation when used by insurgent politicians from the far-ends of the political spectrum. Based on that, we decided to include pejorative and more neutral terms in the elite-dictionary as well as general and more specific terms if the latter stood, pars pro toto, for a broader category of elites. An analogous argument can be made for the salience of ‘people’. Here, all terms that referred to ‘people’ as a political community, or a community of fate subjected to the same conditions were retained. The people vocabulary was less varied than the elite vocabulary, with a few terms making up for almost all mentions (e.g. people, Americans, peuple, Français, compatriotes). The pronouns ‘we’ and ‘nous’ and the words ‘America’ and ‘France’ were left out.

Dictionaries were validated through reliability tests with human coders on a subset of the material (1500 tweets, 4500 judgements through Crowd Flower, an online crowd-worker platform, see Lind et al., 2017). Final reliability was assessed for each of the five custom dictionaries through approximate percent agreement (>80%), Cohen’s Kappa (>.60), as well as precision (>.60) and recall (>.80) (See Note 2).

Next, sentiment was scored using the Lexicoder Sentiment Dictionary (LSD), a set of word lists that combine several existing dictionaries (Young and Soroka, 2012). The added benefit of the LSD versus other, similar word lists is that the LSD is available in French (Duval and Pétry, 2016) and English. The authors added to existing LSD lists to include terms directly related to negative or positive valence in tweets from the corpus. While unable to capture the full meaning of a discourse, this approach enables researchers to gauge the recurrence and tonality of specific topics and targets in candidate messages.

The unit of analysis for word scoring is the 140-character text at the tweet level. Each tweet represents one unit of analysis in the sample, and comments attached to tweets are not analysed. Thus, the corpus is drawn from the candidate’s text only. The dictionaries were divided in five sub-categories: (1) political elite, which includes references to politicians as a category, specific political actors that were often targeted and used as symbols, and institutions; (2) corporate elite including standard terms referring to major economic forces, such as corporations, and ‘the wealthy’; (3) the media elite consisting of references to leading news organizations, journalists, and media managers; (4) the people including all expressions referring to the people as a socio-political category and (5) immigration including a list of words related to this topic.³

Results

Targets

The first hypothesis (H1) proposed that right-leaning conservative candidates talk about political more often than economic elites and institutions. Table 1 reports the frequencies and percentages of tweets with at least one reference to ‘the people’ and ‘the elite’, broken down by elite category. About 28% of Trump’s tweets include at least one political elite mention and about 15% of Le Pen’s tweets mention political elites. In line with H1, both right-leaning candidates talk more about political than economic elites compared to their opponents. H2 expects that left-leaning candidates mention economic elites more often than political elites. According to Table 1, left and liberal candidates make the political establishment less salient in their discourse. Sanders mentions the corporate elite most often across the sample (28% of the time), and Clinton does so almost as often (24% of the time) which gives her discourse a left-wing touch. In contrast, Trump rarely refers to corporations or issues of economic class. Instead, he regularly invokes the media (18% of the time), more so than any other candidate in the sample. However, Sanders and Le Pen refer to the people more often than their opponents. About 18% of Sanders’s tweets include ‘people’ references, and 12% of Le Pen’s do. H3 proposed that right-leaning candidates would mention immigrants and borders more often as they define who belongs in the social out-group. Le Pen (10%) and Trump (2.7%) indeed talk about immigration more often than other candidates, both in terms of raw mentions and proportion of all tweets over the year.

As regards the difference between the moderate and the more extreme anti-mainstream candidates (H4), we find that the insurgents Trump, Sanders and Le Pen are more people-centric than Macron, but surprisingly the latter employs more people-centric rhetoric than far-left insurgent Mélenchon. This result, however, is probably due to an undercounting of Mélenchon’s references to ‘the people’ because of an ambiguous vocabulary (he tends to use the word ‘gens’ instead of the word ‘peuple’) that is hard to capture with a populism dictionary. Post hoc Chi-square tests suggest statistically significant differences in how frequently the candidates refer to the categories ‘the elite’, ‘immigration’, and ‘the people’ (Table 1; p > .001). This pattern of mentions reveals Sanders’ discourse as the only left-wing populist, whereas Clinton’s lacks the people-centric component to deserve this label, and Mélenchon does not make economic elites salient. Only Le Pen’s discourse fulfils all requirements of complete populism.

Targets and sentiment

Elite mentions alone do not account unambiguously for oppositional discourse, but they do offer a rough estimate for how candidates construct these categories. The combined analysis of target and sentiment partially address this limitation, proposing that sentiment is related to populist targets in predictable patterns. Table 2 breaks down the relationship between references to populist categories and overall sentiment of a tweet to further explore the ideological coloration of their populist discourse (H1, H2 and H3). Furthermore, the sentiment dictionaries should reveal that negative tone appears alongside mentions of the elite, and positive tone should appear alongside mentions of the people especially for candidates who most fundamentally challenge the establishment from left or right ideological convictions (H5).

Table 2.

Frequency of positive (P) and negative (N) tweets, and net sentiment (Net) by dictionary category and candidate.

	Political elite		Corporate elite		Media elite		Immigration		The people		All tweets
	N/P	Net	N/P	Net	N/P	Net	N/P	Net	N/P	Net	Net
	(%)	M	(%)	M	(%)	M	(%)	M	(%)	M	M
US candidates
Clinton	35/58	.031	54/45	−.015	52/52	.010	54/49	−.015	47/59	.009	.016
Sanders	55/52	−.001^a	65/57	−.005	55/16	−.029	79/55	−.052	62/63	.001	−.006^a
Trump	69/51	−.030^a	57/73	.016	50/67	.017	55/66	.004	31/81	.079^a	.037^a
French candidates
Le Pen	57/60	−.002^a	49/61	.012	31/48	.014	71/40	−.068^a	45/59	.018	−.003^a
Macron	32/53	.020	40/61	.025	16/83	.074	62/59	−.017	40/54	.010	.023
Mélenchon	51/43	−.014^a	33/41	.003	50/51	.006	69/42	−.049	34/43	.006	−.004^a

statistically significant difference (TUKEY) from the reference candidate (Clinton in the United States, Macron in France) at 90% confidence interval.

This is the case for tweets referring to ‘the people’ where all candidates’ net sentiment scores are positive (net sentiment greater than zero). Trump tends to include positive terms when he mentions the people more often than other candidates (81% of the time; (F(5, 1381) = 22.93, p < .001)). Both Sanders and Trump employ negative terms when they mention political elites (55% of Sanders’ tweets that mention political elites also contain negative terms vs. 69% of Trump’s; (F(5, 2613) = 24.39, p < .001)). Likewise, the more extreme outsider French candidates Mélenchon (net sentiment = –.014) and Le Pen (net sentiment = –.002) are also more negative on average when they mention political elites, and their mean net sentiment is statistically more negative than Macron’s (p > .001). In contrast, political elite mentions of the more mainstream candidates Clinton and Macron tend to include positive terms. These results support H1 more clearly in the American context than in the French.

In the corporate elite category (Table 2), Sanders and Clinton are negative on average when they talk about corporate elites, and Trump is positive. The French candidates are more positive on average when they talk about corporations on Twitter, and there is no statistically significant difference among them. It is curious that the same left-right pattern as in the United States did not exist between them. Here, the French version of the sentiment dictionary might not be honed sufficiently to capture all the negative framing of corporations that Le Pen and Mélenchon do employ in certain tweets. Thus, the combined analysis of targets and tone only partially confirms H1 and H2. H3 is also partially supported: In line with her party’s ideology, Le Pen is decisively negative when she talks about immigration (mean net sentiment = –.068), and post hoc tests suggest she is more negative than her French opponents ((F(5, 560) = 7.23, p < .001)). The same cannot be said for Trump though.

Finally, to further address H5, the relationship between sentiment and targets on Twitter was explored in the more stringent multi-level modelling (MLM) framework.⁴ We calculated the odds ratios that a positive or negative term will coincide with a mention of ‘the elite’ (political, corporate, or media elite) or ‘the people’ with all tweets (Figure 1). Negative sentiment is positively related to mentions of the elite (OR = 1.23, p < .001), but there is no statistical relationship between negative sentiment and mentions of the people (OR = 1.03 p = n.s.). That is, for each additional negative word in a tweet, the odds of a tweet also containing a mention of the political elite increases by about one and a quarter, while negative word mentions are not related to references to the ‘people’. In contrast, each additional positive word in a tweet decreases the odds of a tweet also containing mentions of the political elite (OR = .88, p > .001), but each additional positive word count increases the odds a tweet will also mention the people (OR = 1.15, p > .001).

Figure 1.

Random effects (intercept) of the probability (P) that a sentiment keyword coincides with a reference to the people or elites by country and candidate. Sentiment is centred to the grand mean by country.

However, there is considerable variation between candidates in how sentiment is related to mentions of ‘the elite’ (variance = 3.29, Conditional R² = 15%) and ‘the people’ (variance = 3.29, Conditional R² = 10%). Figure 1 plots the probability that a sentiment term coincides with mentions of either of these constructs by country. The centre line represents the grand mean, and candidates can either be more likely, or less likely to employ sentiment relative to their opponents. For example, the probability that Trump will talk about political elites with negative tone (Figure 1, top panel) is about 20% higher than the sample mean, while Clinton is less likely to talk about political elites negatively. Sanders is significantly more likely to talk about the people with a positive tone than the rest. For the French (bottom panel), the probability Le Pen will talk about the people positively and political elites negatively, is about 50% higher than for her opponents. This means that the allegedly most populist and extreme candidates based on their political positions (Trump, Sanders, Le Pen) in both countries are clearly demarcated from their main liberal opponents in terms of the positive framing of the people and the negative framing of the elite, lending strong support for H5.

Discussion

This study analysed variation in tone and targets of populist-style rhetoric in the social media campaigns of insurgent presidential candidates whose campaigns had international impact. It also suggested new modes to measure the presence of populist rhetoric in social media messages. The employed method was able to identify variation within the discourse of the six candidates based on: (a) the number of references to different types of ‘elite’, especially political and corporate segments, (b) the use of sentiment overall and (c) a stronger use of sentiment when talking about ‘the elite’ and ‘the people’. Insurgent candidates from the far-ends of the political spectrum, regardless their ideology, scored much higher in these categories compared to more moderate candidates like Macron and Clinton.

On the thematic (targets) and the affective (tone) level, we distinguish between discourses that share barely a trace of populism, epitomized by the moderate and liberal candidate Macron, and a clearly populist rhetoric espoused by Le Pen, Sanders and to a lesser extent Trump. The congruence between our classification and previous research (Gerstlé and Nai, 2019; Olivier and Rahn, 2016; Zulianello et al., 2018) supports the validity of the computer-aided dictionary method for multilingual research in the quantitative paradigm.

While our results confirm that candidates who are located closer to the far-ends of the political spectrum are more likely to resort to populist rhetoric (which was also found by Ernst et al., 2017), they nevertheless point to ideological nuances perceptible in the targets of a discourse that focuses on the elite and their failures. Although the politicians representing left or right ideologies employed more negative sentiment than the moderate candidates, and were more negative when referring to the political elite, a nuanced typology describes the discursive patterns better than a blunt distinction between populist and non-populist communication patterns.

While Trump and Le Pen both resort to populist rhetoric, Trump’s rhetoric resembles an economically liberal and political-cultural anti-establishment type of populism. In contrast, Le Pen targets all segments of the ruling elite except the media (although, she occasionally does as a qualitative analysis reveals), and references to ‘the people’ play a stronger role in her communication compared to Trump. Her rhetoric is also more geared towards the people as democracy’s sovereign who has somehow been betrayed of their legitimate power. Moreover, an in-depth examination of Le Pen’s tweets reveals a tendency to present the citizen-elite relationship as exploitative, adding a socialist component to her rhetoric to which the quantitative analysis already pointed. Since also she stands out in terms of references to immigration, her Twitter discourse can be characterized as complete populism according to Jagers and Walgrave’s (2007) typology. Le Pen’s rhetoric must also be seen as against the background of her aim to reshape the party’s nationalist image (dédiabolisation). Populism gives her the rhetoric toolbox to bring together identity politics and the defence of the social welfare state in one coherent discourse.

A conclusion beyond the cases analysed concerns the usefulness of the broad distinction between populist and non-populist candidates. Due to the differences even among insurgent candidates, populist communication should be thought of as occurring in ideologically coloured subtypes. This means that researchers should be more specific when they invoke the populist tag. For instance, Sanders’ rhetoric is better characterized as a populist framing of a class conflict, while Le Pen’s rhetoric is directed against the political power system as a whole and has Rousseauian radical-democratic, socialist and nationalist undertones. Trump’s populism is, by contrast, directed against a cultural and political elite but does not question the economic order.

Despite the efficiency of the automated analysis, the measurement instrument has some limitations when candidates refer to ‘the people’ and ‘the elite’ with a specific vernacular. Mélenchon avoids the French term ‘peuple’ and systematically replaces it by the less-charged word ‘gens’, which is more ambiguous and often carries a non-political connotation. Thus, the initial instrument as applied here was undercounting references to ‘the people’ for the French left. A similar problem was found for Trump, whose overall immigration mentions are lower than expected (less than 3% of tweets). In-depth analysis shows that this is because Trump often replaces typical immigration terms (border, immigrant) with ethnic or racial cues which our dictionary did not account for. More generally, a weakness of a dictionary-based approach is that it is based on a limited number of terms that seldom capture the full range of relevant expressions. In addition, our approach is unable to capture deeper layers of meaning that other methods such as critical discourse analysis may be able to analyse better. Moreover, the meaning of some words operationalizing populist references to ‘the elite’ (‘the people’) are ambiguous: They carry a negative (positive) and thus ‘populist’ connotation in some instances but not always. Thus, the dictionary-based approach has some problems with separation between the measurement of an issue focus and blunt anti-elitist or pro-people claims. Nevertheless, we would argue that these word lists are better at parsing general patterns, while the more precise conclusions should come from careful in-depth analysis.

Despite the limitations, capturing the nuance in this type of rhetoric can already guide future studies in identifying the specific value systems that drive anti-elite discourse. A rough populist versus non-populist distinction may obscures these divergences. That is, it is difficult to predict the real-world implications of populist language without an accounting of the ideological tinges.

Supplemental Material

elite_en1307 – Supplemental material for What kind of populism? Tone and targets in the Twitter discourse of French and American presidential candidates

Supplemental material, elite_en1307 for What kind of populism? Tone and targets in the Twitter discourse of French and American presidential candidates by Peter Maurer and Trevor Diehl in European Journal of Communication

Footnotes

Funding

The author(s) received no financial support for the research,authorship,and/or publication of this article.

ORCID iD

Peter Maurer

Supplemental material

Supplemental material for this article is available online.

References

Bonikowski

Gidron

(2016) The populist style in American politics: Presidential campaign discourse, 1952–1996. Social Forces 94(4): 1593–1621.

Borra

Rieder

(2014) Programmed method: Developing a toolset for capturing and analyzing tweets. Aslib Journal of Information Management 66(3): 262–278.

Bracciale

Martella

(2017) Define the populist political communication style: The case of Italian political leaders on Twitter. Information, Communication & Society 20(9): 1310–1329.

Bruns

(2012) How long is a tweet? Mapping dynamic conversation networks on Twitter using Gawk and Gephi. Information, Communication & Society 15(9): 1323–1351.

Canovan

(1999) Trust the people! Populism and the two faces of democracy. Political Studies 47(1): 2–16.

De Vreese

Esser

Aalberg

, et al. (2018) Populism as an expression of political communication content and style: A new perspective. The International Journal of Press/Politics 23: 423–438.

Diehl

Vonbun-Feldbauer

Barnidge

. (2019). Tabloid news, anti-immigration attitudes, and support for right-wing populist parties. Communication and the Public Adcanced online access.

Duval

Pétry

(2016) L’analyse automatisée du ton médiatique: construction et utilisation de la version française du Lexicoder Sentiment Dictionary. [The automated analysis of media tone: Construction and utilization of the French version of Lexicoder Sentiment Dictionary]. Canadian Journal of Political Science 49(2): 197–220.

Engesser

Ernst

Esser

, et al. (2017) Populism and social media: How politicians spread a fragmented ideology. Information, Communication & Society 20(8): 1109–1126.

10.

Ernst

Engesser

Büchel

, et al. (2017) Extreme parties and populism: An analysis of Facebook and Twitter across six countries. Information, Communication & Society 20(9): 1347–1364.

11.

Ernst

Blassnig

Engesser

Büchel

Esser

(2019) Populists prefer social media over talk shows: An analysis of populist messages and stylistic elements across six countries. Social Media + Society. https://doi.org/10.1177/2056305118823358

12.

Escalona

(2017) Le «moment Macron» [The ‘Macron moment’]. Cités 72(4): 175–185.

13.

Gerbaudo

(2018) Social media and populism: An elective affinity? Media, Culture & Society 40(5): 745–753.

14.

Gerstlé

Nai

(2019) Negativity, emotionality and populist rhetoric in election campaigns worldwide, and their effects on media attention and electoral success. European Journal of Communication 34: 410–444.

15.

Gidron

Bonikowski

(2013) Varieties of populism: Literature review and research agenda. Weatherhead Center for International Affairs, Working Paper Series, No.13-0004, Harvard University.

16.

Grimmer

Stewart

(2013) Text as data: The promise and pitfalls of automatic content analysis methods for political texts. Political Analysis 21(3): 267–297.

17.

Hameleers

(2018) A typology of populism: Toward a revised theoretical framework on the sender side and receiver side of communication. International Journal of Communication 12: 2171–2190.

18.

Hameleers

Bos

de Vreese

(2017) ‘They did it’: The effects of emotionalized blame attribution in populist communication. Communication Research 44(6): 870–900.

19.

Hawkins

Rovira Kaltwasser

(2017) What the ideational study of populism can teach us, and what it can’t. Swiss Political Science Review 23(4): 526–542.

20.

Jagers

Walgrave

(2007) Populism as political communication style: An empirical study of political parties’ discourse in Belgium. European Journal of Political Research 46(3): 319–345.

21.

Kaid

Gerstlé

Sanders

(1991) Constructing a political communication project in two cultures. In: Kaid

Gerstlé

Sanders

(eds) Mediated Politics in Two Cultures. Presidential Campaigning in the United States and France. New York: Praeger, pp. 3–8.

22.

Kazin

(2016) How can Donald Trump and Bernie Sanders both be ‘populist’? The New York Times Magazine, 22 March. Available at: https://www.nytimes.com/2016/03/27/magazine/how-can-donald-trump-and-bernie-sanders-both-be-populist.html.

23.

KhosraviNik

(2018) Social media techno-discursive design, affective communication and contemporary politics. Fudan Journal of the Humanities and Social Sciences 11: 427–442.

24.

Laclau

(1979) Politics and Ideology in Marxist Theory: Capitalism – Fascism – Populism. London: Verso.

25.

Lewis

Zamith

Hermida

(2013) Content analysis in an era of big data: A hybrid approach to computational and manual methods. Journal of Broadcasting & Electronic Media 57(1): 34–52.

26.

Lind

Gruber

Boomgaarden

(2017) Content analysis by the crowd: Assessing the usability of crowdsourcing for coding latent constructs. Communication Methods and Measures 11(3): 191–209.

27.

McGregor

Mourão

Molyneux

(2017) Twitter as a tool for and object of political and electoral activity: Considering electoral context and variance among actors. Journal of Information Technology & Politics 14(2): 154–167.

28.

Mény

Surel

(2001) The constitutive ambiguity of populism. In: Mény

Surel

(eds) Democracies and the Populist Challenge. Basingstoke: Palgrave, pp. 1–21.

29.

Mercier

(2016) Twitter as a counter public sphere. Polemics in the Twittersphere during French electoral campaigns. In: Frame

Brachotte

(eds) Citizen Participation and Political Participation in a Digital World. London: Routledge, pp. 139–152.

30.

Mudde

(2004) The populist zeitgeist. Government and Opposition 39(4): 541–563.

31.

Oliver

Rahn

(2016) Rise of the Trumpenvolk: Populism in the 2016 Election. The ANNALS of the American Academy of Political and Social Science 66: 189–206.

32.

Reynié

(2011) Le tournant ethno-socialiste du Front national [The ethno-socialist turn of Front national]. Études 41(5): 463–472.

33.

Rooduijn

Pauwels

(2011) Measuring populism: Comparing two methods of content analysis. West European Politics 34(6): 1272–1283.

34.

Stanley

(2008) The thin ideology of populism. Journal of political ideologies, 13(1): 95-110.

35.

Stanyer

Salgado

Strömbäck

(2016) Populist actors as communicators or political actors as populist communicators. In: Aalberg

Esser

Reinemann

, et al. (eds) Populist Political Communication in Europe. London: Routledge, pp. 353–364.

36.

Stockemer

Barisione

(2017) The ‘new’ discourse of the Front National under Marine Le Pen: A slight change with a big impact. European Journal of Communication 32(2): 100–115.

37.

LYF

Cacciatore

Liang

, et al. (2017) Analyzing public sentiments online: Combining human- and computer-based content analysis. Information, Communication & Society 20(3): 406–427.

38.

Taggart

(2004) Populism and representative politics in contemporary Europe. Journal of Political Ideologies 9(3): 269–288.

39.

Vergeer

Hermans

(2013) Campaigning on Twitter: Microblogging and online social networking as campaign tools in the 2010 general elections in the Netherlands. Journal of Computer-Mediated Communication 18(4): 399–419.

40.

Wells

Van Thomme

Maurer

, et al. (2016) Coproduction or cooptation? Real-time spin and social media response during the 2012 French and US presidential debates. French Politics 14(2): 206–233.

41.

Wodak

(2015) The Politics of Fear: What Right-Wing Populist Discourses Mean. London: SAGE.

42.

Young

Soroka

(2012) Sentimentive news: The automated coding of sentiment in political texts. Political Communication 29(2): 205–231.

43.

Zulianello

Albertini

Ceccobelli

(2018) A populist zeitgeist? The communication strategies of Western and Latin American political leaders on Facebook. The International Journal of Press/Politics 23(4): 439–457.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB