Blaming it on the Elites: A Cross-Country Analysis of Right-Wing Populist Rhetoric on Social Media during the Corona Crisis
Luca Polizzi Student No. 12255289
Master’s Thesis
Graduate School of Communication
Erasmus Mundus Master’s Joint Degree Journalism, Media and Globalisation
Supervised by Sjifra de Leeuw Date of Completion: 29/05/2020
Abstract
Populism and crisis are contested concepts in the field of political communication. This thesis in particular looks at how elite responses to crisis affect the anti-elitist rhetoric of populist politicians. Thus far, conducting such study has been difficult due to the fact that crises have very different characteristics across countries and cases, thereby limiting the opportunity for clean comparisons. The present study uses the recent outbreak of the coronavirus to develop such a clean comparison. Holding the characteristics of the crisis constant across countries, the study looks at how elite responses to the crisis affect anti-elitist rhetoric of right-wing populist leaders in Italy, the UK, Germany, Finland and Sweden on Facebook. The study found no sufficient evidence to claim that there is an effect of the outbreak. There is no statistically significant evidence to state that the politicians increased their anti-elitist rhetoric after the beginning of the crisis or that countries with late responses will have more anti-elitist references than countries with early responses after the beginning of the crisis. However, whilst the predicted effect is not statistically significant, descriptive overviews of our data show that for countries with late responses, anti-elitist rhetoric has increased after lockdown more than it has for countries with early responses. The study concludes with an overview of the findings as well as some considerations on the limitations of the research and some recommendations on ways to expand its focus.
Introduction
The era of social media has brought forward new ways for politicians to communicate with the electorate that remove news media and television filters. In other words, politicians now have more direct means of communication with the people. Inversely, if we want to follow a politician, we can simply like their page and be up to date on what they have to say. It is in fact partly through social media that populism, in particular right-wing populism, has found fertile ground across Europe, becoming a hot topic for discussion within and beyond academia (Kaltwasser, Taggart, Espejo, Ostiguy & Manucci 2017; Hameleers, 2019). Social media offer a space for an affective rather than rational reasoning that is constructed by populist politicians, who use the online tool to boost their electoral support. Populist parties now have more ways to appeal to people’s fears and frustrations towards the government and society to drive electoral support for divisive politics. In effect, it is mainly through social media that right-wing populist politicians communicate their message and mobilise their supporters (Flew & Iosifidis, 2020).
According to Laclau (2005), populism is also linked to crises. He argues that the presence of a crisis is a necessary precondition for the spread of populism. Crises weaken the credibility of the establishment’s rhetoric, which is why the type of rhetoric spread by opposing parties (i.e populist parties) is important. Anti-establishment discourse can reinforce the loss of credibility of the establishment, meaning that crises represent an opportunity for the flourishing of populism. A gap in literature exists when it comes to how elite responses to crises affect populist discourse during a crisis. This is especially important because if anti-elitism is a characterising element of populist discourse, it means that elite responses to crises are part of populist communication and should thus be explored further in academic literature. One of the reasons for this gap in literature is that the nature of crises varies substantially across situations, which limits the comparability between each case. In this thesis, I use the international coronavirus outbreak to study how elite responses affect populist discourse, which allows to overcome the issue of case diversity because all countries endured a similar crisis.
This study therefore seeks to answer the following research question and sub-question: “How does a crisis affect populist discourse? And to what degree is this steered by governments’ responses to this crisis?” The aim of this paper is to offer empirical insights into right-wing populist communication in times of crises and to provide an understanding of the extent to which right-wing populist rhetoric varies in function of the promptness of their government’s response to a crisis. In particular, I argue that populist politicians from countries with delayed responses to the outbreak of the crisis will have a more ‘anti-elitist’ rhetoric than politicians from countries with a prompt response to the crisis.
Existing literature has mostly dealt with the subject of populism and the opportunities created by social media (Kaltwasser, Taggart, Espejo, Ostiguy & Manucci 2017; Hameleers, 2019). While these views can provide valuable insights, they cannot be used to reflect the right-wing populist rhetoric during a pandemic. Such an unprecedented and recent pandemic is still uncharted territory and can provide additional insights into our understanding of populism in times of crisis. In this thesis, I borrow insights from a relatively novel subfield of research on medical populism and apply it to the thus far unconnected research on populist communication on social media.
This study may provide insights for further research focusing on populism in times of crisis generally, as well as the developments of the political phenomenon during and in the aftermath of the coronavirus outbreak. Perhaps most importantly, this research can be used as a starting point to broaden and deepen the analysis, thus including a larger amount of countries as well as more characterising elements of populism in the analysis. From a societal perspective, this research may also provide an initial understanding of whether or not the coronavirus will mark the decline of right-wing populist parties.
Methodologically, I develop a unique test to study the relation between crisis, populist communication and elite responses. I do so by focusing on the international outbreak of the coronavirus, thereby holding the characteristics of the crisis constant across cases. A crisis like the coronavirus outbreak has opened new angles of debate on populism and the effects that the crisis is
having on populist politicians and their parties (Champion, 2020; Fieschi, 2020; Gostoli, 2020; Mudde, 2020; Scott, 2020; Wright & Campbell, 2020). Looking at how right-wing populist politicians in Europe have been communicating the crisis to the public through their Facebook pages can shed light into the subject. It can, in fact, provide an in-depth understanding of how populist rhetoric differs across the politicians and the countries affected by the pandemic and offer insights into how responses to crises trigger increases in anti-establishment rhetoric from populist politicians attempting to seize the situation for electoral gains.
The focus in this thesis, will be on five politicians from different European countries, namely Sweden, a country that has not shown particular responses to the pandemic outbreak, Italy and the UK, understood as countries having had a delayed response to the crisis, Germany and Finland, understood as countries having had a prompt response to the crisis (Erdbrink & Anderson, 2020; Fund & Hay, 2020; Hill, 2020; McCann, Popovich & Wu, 2020; Wittenberg-Cox, 2020). The argument put forth is that politicians from countries with delayed responses to the outbreak of the pandemic will have a more ‘anti-elitist’ rhetoric than politicians from countries with a prompt response to the crisis.
Theoretical Framework
Populist communication, anti-elitism and social media
An extensive body of literature has dealt with the subject of right-wing populism and its rhetoric (Bobba, 2018; Lazardis & Campani, 2017; Postill, 2018; Salgado, 2018). Even though the provided definitions differ to varying degrees, there seems to be relative agreement on what its defining principles are. In this thesis, I focus on one agreed upon characteristic of populist discourse, namely, anti-elitism (Bobba, 2018). Bobba recognises other defining elements of populism such as people-centrism and the denigration of the ‘other’ group. However, considering the limited scope of this paper, the focus will be on anti-elitism only, which is of particular relevance here given the elites’ involvement in situations of crisis, subject that will be discussed into further detail later on.
Over the past few years, populism and populist parties have had unprecedented success across Western countries with support rates doubling since the 1960s (Inglehart & Norris, 2016). Populist parties are now more than ever, becoming the main driving force of divisive and anti-elitist politics and are exerting leverage on people’s fears, doubts and frustrations to target elites and boost their own electoral support. The rise of social media has marked a new era of communication between politicians, particularly populist politicians and the electorate. Social media have made it possible for the anti-elitist messages of populists to reach the electorate without being filtered by news media and television (Bennett and Pfetsch, 2018).
In other words, communication between politicians and people is now more direct, and this has worked as a facilitator for the spread of populism given that social media are the main means of populists to spread their message, particularly their critiques of the governing elites (Kaltwasser, Taggart, Espejo, Ostiguy & Manucci 2017; Hameleers, 2019). Through social media channels, interactivity between users and between users and politicians has also increased significantly and populist leaders also become the main source of information of their support basis. In simple terms, there seems to be a somewhat ‘symbiotic’ relationship between social media and populism that has also facilitated the formation of a dynamic by which populist politicians spark affective rather than rational reasoning. It is through such dynamic of communication that populists manage to spread their anti-elitist message. Instead of providing voters with the tools to be properly informed and critical of political matters, populist politicians frame their message in such a way that they direct popular anger and frustration towards the establishment, blame them for the country’s social and economic issues, and portray them as responsible for not realising the will of the majority of people. This is the basis on which populist politicians structure electoral campaigns and gather votes. They promise to fix the wrong doings of the elites and to fulfil the will of majorities, often to the detriment of minorities. What is most concerning about the political phenomenon is that populists are opposed to liberal democratic values (Arditi, 2007; Lipset, 1960; Moffitt, 2016a; Taggart 2002). This does not mean
that populism rejects democracy. While populists embrace popular sovereignty and the democratic process, they claim that liberal norms and policies are actually harmful and damaging to people. In other words, they present themselves as being opposed to the institutions as well as the economic and social procedures that prevent the realisation of the will of majorities. It is for this reason that they adopt anti-elitist stances, in line with one of the characterising rhetorical elements of populism discussed above. Anti-elitism in fact, is a particularly problematic feature of populist communication because it resonates with the so-called ‘losers of globalisation’, who wish to see the collapse of the establishment, which they hope will result in an improvement of their social status (Petersen, Osmundsen & Arceneux, 2020). These are the voters with which the ‘anti-elitist’ character of right-wing populist rhetoric resonates the most. By blaming the elites for societal issues and by expressing a desire to replace elite-serving institutions with institutions that actually serve the people, populists represent an opportunity for these voters to overthrow the establishment. Anti-elitism is therefore a unifying characteristic of populism, that makes the political phenomenon attractive to a large section of the population.
Populist responses to crises
While the importance of the relationship between populism, anti-elitism and social media has been discussed, it is also relevant to delve into an analysis of the relationship between populism and crises. According to Laclau (2005), crises offer some kind of opening favouring the spread of populism because they weaken the credibility of the establishment’s rhetoric. If that is the case though, why should we study the effects of elite responses to crises on populist politicians? It is actually the type of discourse of populists that favours their flourishment. Anti-establishment rhetoric allows them to reinforce the message that the elites have failed and populist politicians are the best alternative. In other words, when crises occur and governing elites fail in their management, populists can be expected to intensify their anti-elitist rhetoric with the aim of attracting voters and convincing them that populist leaders can do better if elected.
Against this background, it is unsurprising that populist communication flourishes in times of crisis. In some cases, these crises are ‘invented’, in the sense that particular events are exaggerated and presented as a crisis in rhetoric and discourse (Moffitt, 2015). Such invented crises may be related to issues such as immigration, unemployment or social injustices of any kind. It is therefore likely that when an actual crisis presents itself, populists’ will seize the opportunity to critique the establishment. They can claim they can do better than what the elites have done and promise prompt resolutions no matter what the main sources of concern are.
The argument that populists respond to crisis and thrive in their presence is not new. On multiple occasions populist movements gained ground in crisis situations. One example is the environmental crisis, on which right-wing populists such as the Finns party in Finland built an argument of ‘climate hysteria’ (Barry & Lemola, 2019). Another crisis that favoured populists is the 2008 financial crisis, who were also helped by the inability of political leaders to react efficiently and effectively to the crisis (Best, 2018; Moffitt, 2015; Pirro & van Kessel, 2017). Perhaps an even more evident example of a crisis that favoured populists is the migration crisis, which is still ongoing and has contributed to their success across and beyond Europe (Lutz, 2018; Podobnik, Jusup, Kovac & Stanley, 2017; Stojarova, 2018).
In light of this theoretical argument and empirical evidence, it is plausible that: H1. Anti-establishment rhetoric increases after the beginning of a crisis.
The argument can also be taken one step further. If the governing elite has had a delayed and ineffective response or no response at all to the crisis in question, populist politicians will have more reasons and arguments to blame the elite for the crisis and its management as opposed to a government that had an early and efficient response to a crisis. More specifically, while right-wing populist rhetoric has anti-elitist tendencies to start with, one would expect an increase in anti-elitism if a crisis is being mishandled by the establishment. That is based on the argument that if crises are being responded to ineffectively by governments, popular dissatisfaction will grow, particularly when it
comes to economic crises, and populists will seize the opportunity to intensify their anti-elitist rhetoric with the aim of expanding their support basis (Kriesi & Pappas, 2018). Empirical evidence of this can be found in governments’ mishandling of the 2008 financial crisis. The crisis was of such large scale that it caused deep economic and political turmoil, partly due to mistakes of the establishment such as the use of fiscal stimuli to limit the recession (Best, 2018). The unfolding of the events caused increasing dissatisfaction amongst the electorate towards the establishment’s management of the crisis, offering populists an opportunity for expansion.
I therefore expect that:
H2. Anti-establishment rhetoric is even more present in countries that had late responses as opposed to countries with early responses after the beginning of the crisis.
Case selection: The Covid-19 outbreak
So far, there is a lack of cross-country studies linking populism and crisis, more specifically studies looking at the effects of elite responses on populist communication in times of crisis. This is due to the fact that for an accurate study to be conducted, a clean comparison is necessary, meaning that such a topic can be delved into only in the presence of the same crisis across countries or across time. Although most of the large-scale crises this far may have been somewhat similar across countries, such as the 2008 financial crisis and the migration crisis of the past few years, time frames or specific characteristics defining the crises differed, and this limited the comparability potential. It has therefore, not been possible to make comparisons and be sure of whether the effects on populist communication can be attributed to the differences between the crises or elite responses.
The Covid-19 or ‘coronavirus’ outbreak however, represents an opportunity to address this comparability issue given that numerous countries across the world have been enduring the same crisis, characterised by the spread of the pandemic, followed by the introduction of national lockdowns and economic consequences that are yet to emerge on full scale. The pandemic has been
having significant impacts particularly in Europe, which, as already discussed, has a strong right-wing populist presence. Media outlets have in fact, already started discussing the subject extensively, with some journalists touching on the implications the pandemic will have on populism (Champion, 2020; Fieschi, 2020; Gostoli, 2020; Mudde, 2020; Scott, 2020; Wright & Campbell, 2020). Due to the fact that the crisis is of medical nature and its link to populism is the main focus of this study, the relatively novel term ‘medical populism’ comes to mind to refer to the study of populism during medical crises. The term was coined by Lasco and Curato (2019) and for the remainder of the thesis, it will be used to describe populist actors’ response to the Corona crisis.
My case selection was driven by several considerations. I first considered the leaders of the most prominent populist parties across the European countries affected by the pandemic. Then, my case selection was driven in such a way that it ensured sufficient variability in the type of elite responses. In other words, I selected countries that had different responses to the crisis to be able to test my hypotheses. The countries chosen for the analysis are Italy, the UK, Finland, Germany and Sweden. The countries in question were chosen as representative examples of European countries with early responses, delayed responses and countries that chose to not put in place strict measures of containment. The first two can be understood as countries that had delayed responses to the pandemic, Finland and Germany as countries with early responses and Sweden as a country with no response (Helm, Graham-Harrison & McKie, 2020; Migliavacca, 2020; Erdrink & Anderson, 2020; Wittenberg-Cox, 2020). The countries have a strong right-wing populist presence, in Italy being the ‘Lega’ party, in the UK the Brexit Party, formerly known as UKIP, in Germany the Alternative for Deutschland party (AfD), in Finland the Finns Party and in Sweden the Swedish Democrats.
Data and Methods
Data
The social media platform chosen to conduct the research is Facebook. It is one of the largest social media platforms available and it is used by politicians across the world to deliver their message to the
electorate (Larsson, 2015). The five leaders of the above-mentioned parties, namely Matteo Salvini, Nigel Farage, Jorg Meuthen, Jussi Halla-aho and Jimmie Akesson, are all active on the platform with most of them publishing one or two posts a day of varying length except for Salvini, who publishes up to twenty posts every day. Facebook was also chosen because it does not impose a character limit that the politicians have to abide by when posting something, meaning that they are allowed to express their views as extensively as they please. This can perhaps provide additional insights into a study looking at anti-elitism rhetoric. Facebook arguably allows politicians more space to develop and express ideas and critical tones against elites.
The data collection process consisted in gathering all the Facebook posts published by the right-wing populist leaders starting from the date in which the first known case of coronavirus was registered in each country up until the date in which the countries’ governments started easing lockdown restrictions. The first cases of the virus in Sweden were registered on the 31st of January (Sverige Radio, 2020), on the 27th of January in Germany (Berlin Spectator, 2020), on the 29th of January in Italy, the UK and Finland (Il Foglio, 2020; Metro, 2020; Yle, 2020). The total number of posts analysed is 987. The vast majority of the posts was published by Italy’s Matteo Salvini (569). The UK’s Nigel Farage published 169, Finland’s Jussi Halla-aho published 49, Germany’s Jorg Meuthen published 106 and Sweden’s Jimmie Akesson published 95. To be able to test our hypotheses, the date in which countries introduced measures to contain the spread of the pandemic was chosen as the moment marking the ‘beginning of the crisis’ as stated in our hypotheses. The date was selected because the introduction of lockdown measures marks the moment in which popular malcontent started due to financial concerns as well as the moment in which rhetoric on the pandemic itself reached its peak. Restrictions began to be put in place on March 9th in Italy, March 23rd in the UK, March 13th in Germany, March 27th in Sweden and 28th of March in Finland (Horowitz, 2020; Doyle, 2020; Henden, 2020; Teivainen, 2020; Ahlander, Johnson & Pollard, 2020). The end dates of the data collection are April 27th in Italy, April 20th in Germany, and May 14th in Finland (Canettieri & Pirone, 2020; “Coronavirus”, 2020; Government Communications Department, 2020). Data collection was
also ended on May 14th in Sweden and the UK. This date was chosen out of convenience due to time constraints. The UK has in fact announced intentions to ease the lockdown on May 18th, date that goes beyond the scope of this thesis (“Coronavirus”, 2020). As for Sweden, the few measures that have been put in place in the country will be kept until further notice.
Method
Dependent Variable
In this study, I focus on a core element of populism, namely, anti-elitism, which has already been discussed in previous sections. To measure anti-elitist rhetorical elements in social media posts, I first manually detected every reference to opposing political parties, governments, politicians, the EU or critiques of the measures taken to face the crisis using a predefined keyword list (see Appendix). For each hit, I then determined whether the reference was negative or not.
Independent variables
The independent variables in the study are the elite response and the time of response. More specifically, each reference to elitism was coded with ‘0’ or ‘1’ depending on whether the post was published before or after the introduction of lockdown measures, understood as the start of the crisis. The choice was made in light of the fact that the introduction of the lockdown corresponded to the beginning of popular dissatisfaction and criticism towards how the coronavirus situation was being handled. It is also the time in which the discourse focused on the pandemic was at its peak. Elite response is another independent variable that refers to the country of the politician publishing the posts. As already mentioned throughout the paper, Italy and the UK are understood as countries with a late response, while Germany and Finland have had early responses. Sweden was chosen as a country with no response.
Reliability
Before proceeding with the actual coding process, a significant portion of the social media posts had to be translated given my limited understanding of the German, Finnish and Swedish languages. Due to time constraints, contacting coders who could analyse the posts in the three languages was not feasible, which is why Google Translate, a standard online translation device, was used to translate the posts from German, Swedish and Finnish. For this kind of analysis, which only requires coders to identify negative, positive or neutral references to the elites, an advanced knowledge of the languages in question is arguably not required. Nevertheless, it is still important that the measure used is valid and reliable. For this reason, I have undertaken several steps to ensure the reliability and validity of the dependent variable. First, after having collected the data, the codebook was tested through a preliminary analysis of around 10% of the total number of social media posts, so 100 posts. The keyword list was formulated based on this initial analysis of the Facebook posts. To be sure to include the widest variety of potentially anti-elitist references, 20 posts for each politician were used for the preliminary analysis to avoid missing specific terms that certain politicians may not use whilst others do. Following the preliminary analysis, the codebook was updated to improve replicability. An additional test was then conducted through an additional coder. The coder was instructed to analyse the same 100 posts taken from the collected data using the codebook (see Appendix) as a framework to proceed with the coding. I myself recoded those 100 posts, and based on the feedback provided by the coder as well as my own impressions of the analysis, a final revision of the codebook was made. Krippendorf’s alpha was conducted to test inter-coder reliability. The test was conducted for the dependent variable ‘Valence’, coded as a nominal level of measurement. The results (alpha= .798) indicate a fairly reliable level of agreement amongst the coders.
For the research, a quantitative content analysis was deemed the most appropriate method. The advantage of this type of analysis is twofold. First, quantitative content analysis is the most suitable research method for large scale studies, which are necessary to enable researchers to test comparative
hypotheses. The drawback of conducting a quantitative analysis is that the ‘depth’ of the content analysis is limited, in the sense that it focuses on specific terms rather than phrases, sayings, metaphors, text positioning and so on (Talbot, 2007). It may thus overlook potentially relevant linguistic features. However, focusing on specific concepts and phrases expressed in the social media posts would be significantly time-consuming and analysing a large amount of posts would not be feasible. Secondly, a quantitative study allows researchers to reach findings that can be replicated because it requires a more standardised and systematic technique, meaning that it can, with reliable and valid measurements, be reproduced by other researchers (Reinard, 2006).
Analysis strategy
The statistical analysis chosen to test my hypotheses is a difference in difference with my Difference 1 being my independent variable ‘Lockdown’ and my Difference 2 being the difference between my D1 in late response versus no response cases. The Difference in difference analysis is tested through two binary logistic regression models. In order to run the binary logistic regressions, the independent variables have been coded into four dummy variables: the variable ‘Response’ has been recoded into ‘dummy early response’ (with levels 1 for early response, 0 for the other levels) and ‘dummy late response’ (with levels 1 for late response, 0 for the other levels); and the variable ‘Lockdown’ has been recoded into ‘dummy after lockdown’ (with levels 1 for after lockdown, 0 for the other levels).
Model 1 is used to test the main effect of the predictor variable ‘Lockdown’ to find out whether anti-establishment rhetoric increases after the beginning of the crisis, which is in this case understood as the date in which lockdown measures were introduced. Model 2 is used to test whether anti-establishment rhetoric is more present in countries that had late responses to the crisis than it is in countries with delayed responses. In model 2, an interaction effect between ‘Lockdown’ and ‘Response’ was included to test for the moderation and find out whether the predictor variables have an effect on anti-elitist rhetoric (present under DV ‘Valence’).
Results
In this section, the results of the two logistic regression models are presented. Overall, the results show low statistical significance across variables, meaning that there is not sufficient evidence to suggest that the independent variables have an effect on anti-elitist rhetoric.
A consideration to be made is that initially, both Model 1 and 2 included dummy variables linked to the politicians of both countries with early response as well as countries with late response. However, results showed that the politician dummy variables were seemingly collinear with other variables. The models were thus rerun without the dummy variables for the politicians.
Our Model 1 was used to test our first hypothesis (H1), which states that anti-elitism increases after the beginning of a crisis, here understood as the date in which lockdown measures were introduced. First of all, the results of the Omnibus Test in Model 1 do not reach statistical significance (p-value= .646), which suggests a poor fitness of the model. As for the actual results linked to the variables, Table 1 reports our three variables ‘Early Response’, ‘Late Response’ and ‘After Lockdown’. None of our three variables reach statistical significance given their high p-values. Their standard errors (S.E) are also low in relation to the coefficients (B), which is an indicator that the variables will not be statistically significant. In particular, our variable ‘After Lockdown’ (p-value= .220), which is of direct interest to test our first hypothesis, fails to attain statistical significance at 95% level, meaning that we cannot eliminate the possibility that the relationship between anti-establishment rhetoric and the after lockdown is due to chance. Therefore, the null hypothesis cannot be rejected. Low statistical significance also means that a clear conclusion about its coefficient (B= -.151), cannot be drawn (Field, 2009). However, the fact that the coefficient is negative would suggest that it is more likely to encounter non-negative references to elitism after the introduction of lockdown measures. This would mean that not only our H1 is not supported, but that evidence actually suggests tendencies in the opposite direction of our H1. One must keep in mind though, that the results are not statistically significant, so making such claim would be far-fetched.
Table 1: Model 1 Binary Logistic Regression
Model 1
B (S.E) Statistical Significance
Early Response -.017 (.211) .937
Late Response .100 (.226) .658
After Lockdown -.151 (.123) .220
Intercept .879 (.212) .000
N 1,528
The ‘Valence’ variable was coded as 0 (non-negatives) and 1 (negatives).
To go into more detail into the results of our first model, we also calculated the odds of the references to elitism being negative before lockdown as well as the anti-elitist references after lockdown. We then calculated the change in proportion between the two odds. The odds of negative references to elitism before lockdown for no response countries are e^(Intercept), which in our case is 2.41. The odds of anti-elitist references after lockdown for no response countries on the other hand are e^(Intercept + After Lockdown coefficient), which equals to 2.07. We then calculated the proportional change in odds by dividing the odds of anti-elitist references after and before the lockdown, with a result of 0.85. This means that the odds of a reference to elitism being negative after the lockdown is 15% lower than of the references being negative before lockdown. This confirms the conclusions drawn from the analysis of the p-value and coefficient of the ‘After Lockdown’ variable. The results contradict our H1, which is therefore unsupported.
Table 2: Model 2 Binary Logistic Regression
Model 2
B (S.E.) Statistical Significance
Early Response .250 (.331) .449 Late Response .402 (.417) .335 After Lockdown .248 (.405) .541 Interaction After Lockdown*Early Response -.434 (.430) .313 Interaction After Lockdown*Late Response -.455 (.500) .364 Intercept .624 (.320) .051 N 1,528
Our Model 2 (see Table 2) was used to test our second hypothesis (H2), which states that anti-establishment rhetoric is even more present in countries that had late responses as opposed to countries with early responses after the beginning of the crisis. Again, to run the model, the decision was made to remove the dummy variables linked to the politicians due to their collinearity with other variables. The model was thus run with the same independent variables as Model 1 with the addition of the two interaction terms ‘After Lockdown by Early Response’ and ‘After Lockdown by Late Response’, which were needed to directly test our H2. Results are partially similar to the results emerging in the first model. The Omnibus Test shows low statistical significance (p-value= .744),
which suggests a poor fitness of the model. As for our two interaction terms, they both have low statistical significance, with ‘After Lockdown by Early Response’ having a p-value of .313 and ‘After Lockdown by Late Response’ having a p-value of .364. If we were to comment on their effect even though they are not statistically significant, the fact that they have negative coefficients means that in both cases, you are more likely to encounter positive posts both in countries with early responses and countries with late responses after the introduction of lockdown measures in comparison with the country with no response, which was our control variable. Again, the fact that the results are not statistically significant means that we cannot reject the null hypothesis and thus cannot accept the alternative hypothesis.
More in detail into the results of our second model, we calculate the difference in odds between before and after lockdown for both early response and late response countries. The odds of the anti-elitist references before lockdown for early response countries are e^(Intercept + Early Response coefficient), which is 2.4. The odds of references to elitism being negative after the lockdown for early response countries are e^(Intercept + Early Response coefficient + After Lockdown coefficient + Early Response*After Lockdown coefficient), which is 1.99. The proportional change in odds are then calculated, with a result of 0.83. This means that the odds of references to elitism being negative after lockdown are 17% less than of the references being negative before lockdown for early response countries. As for the results for late response countries, the odds for anti-elitist references before lockdown for late response countries are e^(Intercept+ Late Response coefficient), so 2.79. The odds of references to elitism being negative after lockdown for late response countries are e^(Intercept + Late Response coefficient + After Lockdown coefficient + Late Response*After Lockdown coefficient), so 2.27. We then calculate the proportional change in odds with a result of 0.81. This means that the odds of negative references to elitism after lockdown is 19% less than of the negative references before lockdown for late response countries. Based on these results, I find that it is less likely to encounter negative posts after the lockdown than before the lockdown. Actually it seems
that this effect is slightly more present amongst late response countries. Again, this shows a lack of evidence to support our H2.
A relevant additional consideration that has to be made in regards to the results, including the odds calculated in our two models, is that they are predicted results, meaning that they depend on the model, and may thus not actually reflect the presence of anti-elitism within the actual data collected. Therefore, since the poor fitness of the models has already been established, given its overall lack of statistical significance and low classification accuracy, its predicted odds and general results should not be entirely relied upon and do not necessarily represent a valid interpretation of the data.
Robustness Tests
Table 3: Model 1 Binary Logistic Regression without Germany Data
Model 1
B (S.E) Statistical Significance
Early Response -.561 (.259) .030
Late Response .020 (.228) .929
After Lockdown .293 (.190) .124
Intercept .596 (.229) .009
N 673
The ‘Valence’ variable was coded as 0 (non-negatives) and 1 (negatives).
While the results of the two models have already been presented, the first model was tested a second time removing the data linked to the ‘Meuthen’ variable (see Table 3). A consideration that was made in regards to the data is that it represents over half of the entire dataset (see Graph 1), meaning that the large amount of data units pertaining to the variable can affect and therefore shift results excessively. That is particularly the case if we think about the proportionally large amount of negative
references to elites present in the politician’s posts published before the lockdown (see Graph 2). After having tested the model once, the data linked to the ‘Meuthen’ variable was thus filtered out to find out whether or not the variable affected the results.
Graph 1. Graph 2.
The results firstly show a difference in the model fitness. The Omnibus Test results indicate a good fitness of the model (p-value= .001), which is further supported by the low statistical significance of the Hosmer and Lemeshow Test (p-value=.090), used to test poor fitness of the model. In regards to the ‘After Lockdown’ variable, which is the variable is used to directly test our H1, the main difference between the two versions of the first model is that the coefficient linked to the variable is now positive (B= .293). This would suggest that, after removing the data linked to the German politician, we are more likely to encounter anti-elitist references after the lockdown. Although such result would be more in line with our H1, the results linked to this variable do not reach statistical significance. This, again, means that the null hypothesis is not rejected.
Discussion
The aim of this paper has been to look at how crises affect populist discourse and the degree to which such discourse is steered by governments’ responses to crises, which sheds further light into the subject of populism and crisis in political communication. The results emerging from our quantitative content analysis showed that anti-elitism is certainly a rhetorical element that is strongly present in
121 53 335 855 164 0 200 400 600 800 1000 Sweden UK Italy Germany Finland
Number of elite references per country 0.00% 20.00% 40.00% 60.00% 80.00% 100.00% Sweden UK Italy Germany Finland
Percentages of anti-elitist references before and after lockdown
Anti-elitist references before lockdown Anti-elitist references after lockdown
right-wing populist social media discourse across the five European countries (see Graph 3) during the pandemic, which is our selected crisis for the study. However, such element is fragmented and the patterns of presence predicted with our hypotheses were not confirmed. We found no significant confirmation of the effects that elite responses have on populist communication in times of crisis. While the statistical results may suggest that our hypotheses are not validated, it is also important to remember that our models showed poor fitness and low classification accuracy, meaning that they may or may not reflect a valid interpretation of the data. If we look at Graph 4 for instance, it is true that the valence of the elitism element varies in presence depending on the country. It is also true that there does not seem to be a general increase of anti-elitist posts after the beginning of the crisis when it comes to early response countries. However, such increase can be seen in Sweden, the UK and Italy. What this means, is that while the politicians from the countries with late responses, so Italy and the UK, published the largest share of negative posts after the lockdown, the politicians from the countries with early responses, so Germany and Finland, published the largest amount of negative posts before the lockdown. According to the data on the graph therefore, while our H1 remains unsupported, there does seem to be evidence to support our H2. Again, such results cannot be confirmed from a statistical standpoint, given that our logistic regression model results were not statistically significant, but there do seem to be patterns aligning with our second hypothesis. An alternative explanation for the results that could also be possible is a difference in levels of politicisation of this particular crisis across countries. As Pirro, Taggart and van Kessel (2018) suggest, in the past, populists have been selective when it comes to the crises they have politicised for electoral gains. For instance, a populist like Germany’s Jorg Meuthen has tended to be significantly less prone to being critical of the elites after the introduction of lockdown measures. While that may be due to the successful management of the crisis by the elites, it may also be that a decision was made to avoid politicising this particular crisis considering its sensitivity and the high number of deaths it has caused worldwide. A similar consideration can be made for Finland’s Jussi
Halla-aho, whose Facebook posts tended to decrease in their mentions of elites generally, after the beginning of the crisis, which in our case is after the introduction of the lockdown. Another argument that can be made based on the data present in Graph 4, is that aside from the early response countries, which seem to have changed both the amount as well as the proportion between negative and non-negative references to elitism after the beginning of the lockdown, the other countries did not. This may mean that while in some cases, the crisis has changed the anti-elitist rhetoric of populists, in others, particularly in late or no response countries, it has not. This finding would partially be in line with the conclusions drawn by Pirro and van Kessel (2017) who argue that the crisis they selected for their study had different effects depending on the country in question.
Graph 3. Graph 4.
In general, several considerations can be made as to why our results were not statistically significant. Firstly, the fact that there was such a high number of negative references compared to non-negative references to elitism supports the claim that anti-elitism is a predominant rhetorical element of populist leaders. However, the predominance of negative references to elitism also caused an
0.00% 20.00% 40.00% 60.00% 80.00% 100.00%
Percentage of elitist posts by country Negative Non-Negative 0.00% 20.00% 40.00% 60.00% 80.00% 100.00%
Sweden UK Italy Germany Finland
Percentage of elitist posts before and after lockdown by country
After Lockdown Non-Negative After Lockdown Negative Before Lockdown Non-Negative Before Lockdown Negative
imbalance of classes, that is particularly linked to the data of Meuthen’s posts, which took up over half of the entire dataset, meaning that the predicted effect of the independent variables on the dependent variable may not have been entirely accurate. Admittedly, a more model-fitting way to analyse rhetoric could have been through Twitter. While coding the Facebook posts of each of the politicians, it was noticed that while some politicians posted more often on the social media platform, others did not post as often but did publish significantly lengthier posts. Thus, having posts as the unit of analysis would have caused an imbalance by which the politicians who published more often would have a much larger number of data units. To overcome such issue, the decision was made to use references to elitism as the unit of analysis. However, this caused an imbalance in the opposite direction: politicians who published less often but had more content, had a larger number of data units. For this reason, Twitter, which has a character limit of 140, would allow a cleaner and more balanced comparison between the posts published by the politicians.
A second reason why the results were not statistically significant may be that our data sample was not large enough (Dahiru, 2008). However, our sample included all of the posts published by each of the politicians from the very beginning of the spread of the virus to the beginning of the easing of lockdown measures. A way to overcome this and to increase the sample size whilst maintaining the time frame relevant to the study could be by extending the time frame to before the beginning of the spread, up until the very end of lockdown in each of the countries. Doing so might also favour a more balanced time frame of analysis between the before and after lockdown periods. One must bear in mind however, that the such a large scale analysis would require automated data collection techniques as opposed to a manual coding technique which was employed in this case.
Despite the fact that the findings are overall inconclusive, the paper has succeeded in providing further insights into existing studies on populism in times of crisis (Kriesi & Pappas, 2015; Laclau, 2005; Moffitt, 2015; Pirro, Taggart & van Kessel, 2018; Pirro & van Kessel, 2017). One of the main considerations emerging from the literature is that up until now, studies focusing on the topic of
populism and crisis through cross-country analyses were lacking due to the impossibility to draw clean comparisons across countries given the different characteristics of crises across countries and across time. More specifically, few academic sources have focused on how the establishment’s responses to crises impact populist discourse (Lisi & Borghetto, 2018), and those that were centred on the subject did not do cross-country analyses. The Corona crisis allowed for a cross-country study in which the crisis component could be held constant across nations, rendering the present study feasible. Other than filling up a gap in literature though, this study also represents a starting point for future research into the subject, which can provide more extensive knowledge on the impact that elite responses to crises in the general sense, have on populist rhetoric.
Having discussed the findings and limitations of the present research, it is also important to discuss ways to expand research on the subject. First, given its limited scope, this study was centred on anti-elitism only, even though the defining characteristics of populism also include people-centrism and the denigration of an ‘other’ group (Albertazzi & McDonnell, 2008; Mudde, 2004; Bobba, 2018; Engesser, Ernst, Esser & Buchel, 2017). Future studies could in fact reproduce the general framework employed in this paper to study the other two characteristics in the context of a crisis like the coronavirus outbreak. Alternatively, such studies could incorporate the other principles of populism into the present study. For instance, media outlets have been speculating on the Corona crisis representing the ‘end of populism’ given the fact that the ‘other’, which is usually framed as the ‘enemy’ is now invisible, in view of the fact that it is pandemic (Champion, 2020). The anti-immigration argument, which is one of the main rhetorical elements of populists would thus not resonate with an electorate focusing on the issues cause by the outbreak of the virus. Further research could touch on the subject and explore the extent to which anti-elitist rhetoric increased to the expense of anti-immigration discourse during the coronavirus outbreak. Secondly, similar studies on this subject could instead do a qualitative analysis of the Facebook posts of populist politicians published during the pandemic. A qualitative analysis could focus on posts in which populists blame the elites for the mismanagement of the Corona crisis specifically, without including posts in which elites are
being blamed for other political matters. However, such analyses would require familiarity with the political issues of the individual countries as well as fluency in the languages of the countries selected. Finally, further studies could broaden as well as deepen the scale of the present research. Other than including the other defining principles of populism, they could also study social media posts published in more European countries with populist presence.
Bibliography
Ahlander, J., Johnson, S. & Pollard, N. (2020, March 27). Sweden bans public gatherings of more than 50 people: PM. Reuters. Retrieved from https://www.reuters.com
Albertazzi, D., & McDonnell, D. (2008). Twenty-first century populism: The spectre of Western European democracy. Basingstoke: Palgrave Macmillan.
Arditi, B. (2007). Politics on the Edges of Liberalism: Difference, Populism, Revolution, Agitation. Ediburgh: Edinburgh University Press.
Barry, E. & Lemola, J. (2019, April 12). The Right’s New Rallying Cry in Finland: ‘Climate Hysteria’. The New York Times. Retrieved from https://www.nytimes.com
Bennett, W. L. & Pfetsch, B. (2018). Rethinking political communication in a time of disrupted public spheres. Journal of Communication, 68(2), 243-253.
Best, J. (2018). How the 2008 financial crisis helped fuel today’s right-wing populism. The Conversation. Retrieved from https://theconversation.com
Bobba, G. (2018). Social media populism: features and ‘likeability’ of Lega Nord communication on Facebook. Media, Culture & Society, 40(5), 745-753.
Canettieri, S. & Pirone, D. (2020, April 24). Coronavirus: fabbriche, bar, ristoranti e palestre: si parte il 27 aprile, chi riapre e chi no. Il Messaggero. Retrieved from https://www.ilmessaggero.it
Champion, M. (2020, March 27). A Virus to Kill Populism, or Make it Stronger. Bloomberg. Retrieved from https://www.bloomberg.com
Coronavirus: First moves to ease NI lockdown can start next week. (2020, May 14) BBC News. Retrieved from https://www.bbc.com
Dahiru, T. (2008). P-Value, A True Test of Statistical Significance? A Cautionary Note. Annals of Ibadan Postgraduate Medicine, 6(1), 21-26.
Doyle, L. (2020, April 21). Lockdown dates: When did lockdown start? How long have we been in lockdown? Express. Retrieved from https://www.express.co.uk
Engesser, S., Ernst, N., Esser, F. & Buchel, F. (2017). Populism and social media: how politicians spread a fragmented ideology. Information, Communication & Society, 20(8), 1109-1126.
Erdbrink, T. & Anderson, C. (2020, April 28). ‘Life Has to Go On’: How Sweden Has Faced the Virus Without a Lockdown. The New York Times. Retrieved from https://www.nytimes.com
Ernst, N., Engesser, S., Esser, F. (2017). Bipolar Populism? The Use of Anti-Elitism and People-Centrism by Swiss Parties on Social Media. Swiss Political Science Review, 23(3), 253-261.
Field, A. (Eds.). (2009). Discovering Statistics Using SPSS. London, UK: SAGE Publications. Fieschi, C. (2020, March 17). Europe’s populists will try to exploit coronavirus. We can stop them.
The Guardian. Retrieved from https://www.theguardian.com
Flew, T. & Iosifidis, P. (2020). Populism, globalisation and social media. The International Communication Gazette, 82(1), 7-25.
Fund, J. & Hay, J. (2020, April 6). Has Sweden Found the Right Solution to the Coronavirus? National Review. Retrieved from https://www.nationalreview.com
Gostoli, Y. (2020, March 3). How European populists are using coronavirus as a political tool. Al Jazeera. Retrieved from https://www.aljazeera.com
Government Communications Department. (2020, May 5). Government decides on plan for hybrid strategy to manage coronavirus crisis and for gradual lifting of restrictions. Finnish Government. Retrieved from https://vnk.fi/en
Hameleers, M. (2019). The populism of online communities: Constructing the boundary between “blameless” people and “culpable” others. Communication, Culture & Critique, 12(1), 147-165.
Helm, T., Graham-Harrison, E. & McKie, R. (2020, April 19). How did Britain get its coronavirus response so wrong? The Guardian. Retrieved from https://www.theguardian.com
Henden, A. (2020, April 17). Germany lockdown: When did Germany go on lockdown? Why is Germany death rate so low? Express. Retrieved from https://www.express.co.uk
Hill, A. (2020, April 9). Two-thirds of public think UK coronavirus response too slow – poll. The Guardian. Retrieved from https://www.theguardian.com
Horowitz, J. (2020, March 9). Italy Announces Restrictions over Entire Country in Attempt to Halt Coronavirus. The New York Times. Retrieved from https://www.nytimes.com
Inglehart, R.F & Norris, P. (2016). Trump, Brexit, and the Rise of Populism: Economic Have-Nots and Cultural Backlash. SSRN, HKS Working Paper No. RWP16-026, 1-52.
Kaltwasser, C.R., Taggart, P., Espejo, P.O., Ostiguy, P. & Manucci, L. (2017). Populism and the Media. In The Oxford Handbook of Populism. Oxford, UK: Oxford University Press.
Kriesi, H. & Pappas, T.S. 2015. European populism in the shadow of the Great Recession. Colchester, UK: ECPR Press.
Laclau, E. (2005). On Populist Reason. London, UK: Verso.
Larsson, A. O. (2015). Comparing to prepare: Suggesting ways to study social media today – and tomorrow. Social Media + Society, 1(1), 1–2.
Lasco, G. & Curato, N. 2019. Medical Populism. Social Science & Medicine, 221, 1-8.
Lipset, S.M. (1960). Political man: the social bases of politics. Garden City, N.Y.: Doubleday. Lisi, M. & Borghetto, E. (2018). Populism, Blame Shifting and the Crisis: Discourse Strategies in Portuguese Political Parties. South European Society and Politics, 23(4), 405-427.
Lutz, P. (2018). Variation in policy success: radical right populism and migration policy. West European Politics, 42(3), 517-544.
McCann, A., Popovich, N. & Wu, J. (2020, April 5). Italy’s Virus Shutdown Came Too Late. What Happens Now? The New York Times. Retrieved from https://www.nytimes.com
Migliavacca, A. (2020, March 18). Il decreto “Cura Italia” è una risposta inadeguata e tardiva al coronavirus. Secolo D’Italia. Retrieved from https://www.secoloditalia.it
Moffitt, B. (2015). How to Perform Crisis: A Model for Understanding the Key Role of Crisis in Contemporary Populism. Government and Opposition, 50(2), 189-217.
Moffitt, B. (2016a). Populism and Democracy. In The Global Rise of Populism. Stanford, US: Stanford University Press.
Moffitt, B. (2016b). The Performative Turn in the Comparative Study of Populism. American Political Science Association Comparative Politics Newsletter, 26(2), 52-57.
Mudde, C. (2004). The Populist Zeitgeist. Government and Opposition, 39(4), 542-563.
Mudde, C. (2020, March 27). Will the coronavirus ‘kill populism’? Don’t count on it. The Guardian. Retrieved from https://www.theguardian.com
Petersen, M.B., Osmundsen, M. & Arceneaux, K. (2020). The “Need for Chaos” and Motivations to Share Hostile Political Rumors. Retrieved from https://osf.io/
Pirro, A.L., Taggart, P. & van Kessel, S. (2018). The populist politics of Euroscepticism in times of crisis: Comparative conclusions. Politics, 38(3), 378–390.
Pirro, A.L. & van Kessel, S. (2017). United in opposition? The populist radical right’s EU-pessimism in times of crisis. Journal of European Integration, 39(4), 405-420.
Postill, J. (2018). Populism and social media: a global perspective. Media, Culture & Society, 40(5), 754-765.
Podobnik, B., Jusup, M., Kovac, D. & Stanley, H.E. (2017). Predicted the Rise of EU Right-Wing Populism in Response to Unbalanced Immigration. Hindawi Complexity, 1-12.
Reinard, J.C. (2006). Using statistics to conduct quantitative research. In Communication research statistics (pp.3-16). Thousand Oaks, CA: Sage Publications
Salgado, S. (2018). Mediated campaigns and populism in Europe. Cham, Switzerland: Palgrave Macmillan.
Scott, A. (2020, March 18). Coronavirus’ next victim: Populism. Politico. Retrieved from https://www.politico.eu
Stojarova, V. (2018). Populist, Radical and Extremist Political Parties in Visegrad countries vis a vis the migration crisis In the name of the people and the nation in Central Europe. Open Political Science, 1, 32-45.
Taggart, P. (2002). Populism and the Pathology of Representative Politics. In Mény, Y. & Surel, Y. Democracies and the Populist Challenge (pp. 62-80). Basingstoke: Palgrave Macmillan.
Talbot, M. (2007). Media Discourse: Representation and Interaction. Edinburgh: Edinburgh University Press.
Teivainen, A. (2020, April 9). Lockdown of Uusimaa unlikely to continue beyond 19 April, says Marin. Helsinki Times. Retrieved from https://www.helsinkitimes.fi
Wittenberg-Cox, A. (2020, April 13). What do countries with the best coronavirus responses have in common? Women Leaders. Forbes. Retrieved from https://www.forbes.com
Wright, T. & Campbell, K. (2020, March 5). The coronavirus is exposing the limits of populism. Brookings. Retrieved from https://www.brookings.edu
Appendix
Codebook
Post ID:
Elitism references are here understood as references to the EU, current national government coalitions as well as specific politicians of the government and other parties (refer to the Keyword List below for specific guidelines). Specific examples of an elitism elements include: “It’s not a joke! According to the PD, food waste is my fault. There is no limit to the worst.” Or “May this event be an eye opener for the federalists who want to move further decision-making power to Brussels.” Or “Very encouraged by the UK negotiating position. Boris is keeping his promises, for now.” Or “Great speech by David Frost in Brussels, the UK Government is sticking to the promises.” Or “From the point of
view of basic Finns, in addition to ideological differences, there is very little confidence in the current leadership of the Coalition Party.”
References to politicians of the party or to the party to which the populist politician being analysed belongs to should not be included in the analysis.
The analysis should only include text published by the politician. No video content or images with incorporated text should be considered for the analysis.
For each elitist reference, the coder must answer the following questions: 1. What politician is the elitist reference from?
0 Matteo Salvini 1 Nigel Farage 2 Jussi Halla-aho
4 Jimmie Akesson 2. What is the valence of the elitist reference?
0 Positive or Neutral 1 Negative
3. Is the elitism element coming from a politician of a country with late response (UK or Italy), early response (Germany or Finland) or no response (Sweden)?
0 Late response 1 Early response
2 No response
4. How many elitism elements of each valence category are present before the beginning of lockdown measures in the country (general restrictions to limit the spread of the pandemic in the case of Sweden) as opposed to after the beginning of lockdown measures?
0 Before 1 After
List of Keywords for Coding
Elite
Establishment Opposition Party
Names of national politicians Names of the parties
Decision-makers Government EU
European Union
EU institutions and governing bodies Names of EU politicians
Left
Policy-makers Brussels
Finland (When analysing posts by the Finnish party leader) Germany (When analysing posts by the German party leader) Italy (When analysing posts by the Italian party leader) UK (When analysing posts by the British party leader)
Sweden (When analysing posts by the Swedish party leader) Minister
Mayor Governor
