The representation of political leaders in times of pandemic crisis. Evidence from 28 European countries

Over the past decades, the processes of “popularization” of politics (i.e., the exploitation of pop culture codes in political communication, see Mazzoleni and Sfardini, 2009; Mazzoleni and Bracciale, 2020) have received increasing attention from scholars and communicators. In this regard, the increasing relevance of Instagram (and its strong visual component) among the most used social networks represents a small revolution in the contemporary political communication landscape. Although presenting characteristics that are not ideal to convey political messages, Instagram soon became one of the main social networks for political communication (Larsson, 2021). The literature on the topic highlights how politicians’ use of the platform is highly personalized and emotionalized: in particular, the communication style focuses on the figure of the party leader, often engaged in private/pleasant activities (Filimonov, 2016; Ekman et al., 2017). The COVID-19 pandemic crisis, however, has changed dramatically both the ways in which politicians communicate their messages and how people perceive those messages. By drawing upon the literature on political communication during crises (see, for instance, Kahn, 2020), this paper aims at investigating the changes that the pandemic has brought to the visual communication strategies of European political leaders on Instagram. In particular, we will focus on patterns of personalization and emotionalization, by investigating how the pandemic waves have changed the choices of presenting the leader’s face or communicating a positive emotionalization. __Data and methods__ Using a dataset including all posts published by over 150 political leaders from the 27 EU countries and UK, from March 1, 2019, to December 31, 2021 (n = 72,865) and employing automatic face and emotional recognition algorithms, we test the differences in the visual content proposed by populist and mainstream political actors. Face recognition algorithms represent a subset of machine learning algorithms and produce information on the images that are submitted to them in a totally automated way. In particular, they are able to extrapolate several pieces of visual information from the images submitted as input, such as the expressed emotions of the subject (Joo and Steinert-Threlkeld, 2018) and the identity of famous/relevant personalities (such as, as in our case, political leaders) represented in the picture. In the paper, we use the Microsoft Azure recognition algorithm to measure the level of personalization (operationalized as a variable indicating the presence/absence of the party leader in the picture) and emotionalization (the presence/absence of positive emotions) in the picture published by each leader’s account. We focus 1) on the effect that the national COVID-19 death toll has on the levels of personalization/emotionalization and 2) on the effect that these characteristics presented in the published images have on the engagement levels (the number of “favorite” reactions – a measure of endorsement of the message – and the number of comments – a measure of the debate around the message). __Preliminary findings__ By performing logistic multilevel regression models (one for each characteristic of the pictures considered), we find that the effect of the death toll (normalized in a 0-1 scale) on personalization and positive emotionalization is negative and significant (see figure 1). Basically, leaders tend to “disappear” and “stay serious” during the worst moments of the pandemic. __Figure 1. Probability of publishing a positively emotionalized/personalized picture (by death toll)__ __Figure 1 URL: https://ibb.co/0QCyDJt__ Also interesting is the effect that positive emotionalization/personalization have on the engagement of the posts (in this case we employ two models, one having the -logged – number “favorites” as dependent variable and another one doing the same for comments). As far as the “favorite” model is concerned, we do not see any significant change in the aggregate level of endorsement, irrespective of whether the image contains positive emotionalization/personalization. This result seems to hold also if we interact the two characteristics with the national death toll. A different situation is represented in the level of debate around the post (measured with the logged number of comments). A “happy face” or the mere presence of the leader in the picture published increases significantly the number of comments, and this is especially true in situations in which the pandemic is hitting harsher. This might suggest that employing personalized/positive messages during the most dramatic moments of the pandemic might increase critical comments toward the leader. Our preliminary results thus point out that strategies of communicators have been strongly influenced by the pandemic, and that changing those strategies has also changed quantitatively the responses by the general audience.