From Kilowatts to Sovereignty: The Securitization of Energy Discourse in Polish Pre-Election Media (2015–2023)

This article examines how energy security was framed in Polish online media during the parliamentary election campaigns of 2015, 2019, and 2023. Drawing on a mixed-methods design that integrates Structural Topic Modelling (STM) and Critical Discourse Analysis (CDA), it analyses a corpus of 416 articles from sixteen outlets spanning four ideological categories: state-affiliated, PiS-aligned private, liberal-critical, and centrist commercial. The STM identified thirteen latent topics structuring the corpus, and a simulation-based effect estimation procedure revealed 63 credible outlet-topic associations, of which not all are detectable through standard OLS significance testing alone. Three principal findings emerge. First, Polish energy discourse is substantially broader than the coal-versus-gas binary assumed in prior scholarship, encompassing partisan campaign communication, corporate governance, and electoral mobilisation genres alongside substantive policy frames. Second, systematic political parallelism operates at both the distributional and discursive levels: state-affiliated outlets amplified gas independence narratives while liberal-critical outlets concentrated on coal transition challenges, and the CDA demonstrates that convergence in coverage volume masked incompatible problem definitions, causal attributions, and evaluative frames constructed around the same events. Third, a discursive substitution occurred across the three cycles, as the national sovereignty frame displaced the multilateral EU cooperation frame, a shift driven by state-affiliated outlets specifically during the Law and Justice (PiS) governing period. The article advances securitisation theory by demonstrating that energy securitisation in a polarised media system is graduated, ideologically distributed, and contested. It extends Hallin and Mancini’s political parallelism framework from structural alignment to discursive mechanism, and it introduces a four-case inferential reporting protocol for STM effect estimation that can inform future computational social science research.