Comparative computational analysis of the populist discourse of Lithuanian political parties in manifestos and media since 1990

It has been argued that populism is not an ideology, but merely a discursive frame. As such, the analysis of populism would benefit from shifting away from the analysis of the speech of ex-ante defined populist actors to the analysis of political discourse in general. Even though numerous contributions to the literature have done precisely that, research findings are insufficient due to the sample size of discourse and the focus on a limited set of actors in a short time frame. We aim to compare the populism levels of political parties consistently over the years to answer some key questions: 1. How widely political parties employ populist discourse? 2. How does the use of populist discourse change over time? Finally, we ask whether an actor-based distinction between populists and non-populists is possible and relevant in Central and Eastern Europe, which only recently transitioned into democracy. In this paper, we present a new approach to study populism as a discourse based on new natural language processing (NLP) techniques. We assemble a database of near-ideal-type populist texts and from them extract a set of text-features associated with the ideal type populist discourse. We then proceed to measure quantitatively the prevalence of these populist traits in the texts of political parties. We have selected Lithuania during the period 1992-2020 as a case study. Our sample consists of party-electoral manifestos and opinions authored by national-level politicians in the largest online news portal “Delfi.lt” since 2005.