Role of media in fueling affective polarization analysis of online news about Turkey's opposition party

Today, amid the absence of major election-day frauds and overtly violent repression, autocratizing incumbents depend on some level of genuine popular support for political survival. Affective polarization and negative partisanship are considered crucial factors for this support. This thesis attempts to contribute to this line of research by tracking the evolution of content in the media that could fuel this polarization in the Turkish context. It analyzes more than 45.000 online news articles about the main opposition party published since 2009 in two newspapers. Focusing on news from pro-government Sabah and centrist Hürriyet, machine learning algorithms predict the amount of language that portrays opposition as an enemy or rival within politics. Together with the change in ownership of centrist outlet to a pro-government conglomerate in 2018, and the establishment of hyper-presidential system in Turkey, the analysis establishes the long-term trajectory of such language, its changes during election periods, and how it is affected by these junctures across the newspapers. Findings show that vilifying language increases during most election periods for both newspapers, that the pro-government newspaper uses more vilifying language, and that both newspapers got more vilifying towards the opposition after Turkish regime change. In addition, with one newspaper increasingly vilifying while the other reducing the overall coverage of opposition after its change in ownership, it also shows that the influence of incumbent-control over different friendly news outlets may be heterogeneous, depending on the audience.