Avoiding the elephant in the room: The (de-)politicization of Covid-19 and perceptions of political actors in the German Twittersphere

During the 2021 German election, Covid-19 was extremely salient in the public discourse and electorally decisive. Despite its pressing character, however, most parties were attested a depoliticization strategy to avoid electoral losses. While this may facilitate coalition-building after the election, not attending an issue of high public salience can hardly reconcile voters and parties. In contrast, such a strategy raises questions on its consequences for the relationship between political actors and citizens in a non-majoritarian democracy, such as Germany. Against this background, the paper revisits German-language Twitter chatter on Covid-19 four weeks before and four weeks after the election campaign. Our analysis contains two main steps: First, we cluster the conversations into distinct groups and assess the extent to which parties are represented, participate, and mentioned in each cluster on a weekly basis. Thereby, we can estimate the relative levels of depoliticization in Covid-19 talk. Second, we compare how parties are perceived in each cluster and whether there is a relationship between stances addressing political actors and their level of involvement in the conversations. For this purpose, we constructed retweet-based networks based on 8,629,236 Covid-19-related (re-)tweets in German and found eight clearly separable clusters on a continuum either favoring measures or being skeptical of the virus altogether. To measure politicization and stances, we use three approaches: 1) the fraction of party accounts in each cluster, 2) the fraction of tweets authored by party accounts in each cluster, and 3) the frequency with which each party is mentioned in the clusters. For the fraction of party accounts and the fraction of tweets authored by parties, we compiled a list of Twitter accounts affiliated with the seven main German parties. This list contains the Twitter accounts of members of parliament, election candidates and official party accounts (both federal and state accounts) and comprised a total of 1,650 accounts. To count the frequency with which parties were mentioned within the clusters, we extended the list of party-affiliated accounts with names, synonyms, abbreviations and hashtags referring to a person, party or potential coalition and associated them with the specific party they belong to or address. We then conducted a supervised Stance Detection with a pre-trained and fine-tuned RoBERTa-XL language model to infer whether a tweet mentions a political actor with a positive, neutral, or a negative tone. Preliminary findings reveal that clusters shrink in size, particularly after election day. Overall, no one party represents 1% of all accounts or tweets published in a cluster. While politicization is generally low, it is highest in a cluster featuring both conservative and liberal programmatic stances on the handling of the pandemic. Thereby, the Christian Democrats are driving politicization. As governing party providing the Chancellor and the Minister of Health, the CDU could or did not want to escape public dialogue despite an overall preference for depoliticization. Strikingly, the radical-right AfD, which was the only party campaigning on the issue, was comparatively reticent in Covid-19 chatter.