Troublemakers in the Streets? A Framing Analysis of Newspaper Coverage of Protests in the UK, 1992ꟷ2016

In this study, I answer the question: how, if at all, has the way in which mainstream newsmedia frame protest changed over time. I do so using a large-scale newspaper dataset and bycombining a best-practice manual coding technique for frame discovery with different machinelearning algorithms. The study provides new insights into the working of the media on a largeand broad scale while also applying an innovative text-as-data method to a concept which isnotably rare in computational text analysis: framing. Former research indicates that when mainstream news media decide to report about demonstrations at all, protesters often face de-legitimising coverage from traditional news outlets. This phenomenon, dubbed “(journalistic) protest paradigm”, is thought to be a default mindsetthat leads journalists to emphasizes, for example, disruption caused by protesters but to neglectthe aims and arguments behind a protest. If this was true, it would render protest mostly ineffective, as the media is usually the only way in which the majority of citizens can perceive theissues and criticism voiced by a protest. However, a long-term study of protest coverage, which takes into account not just protestsregarding single issues or events but captures a broader picture, is missing from the literatureso far. Likely, this is due to the challenges that come with the size and scope of such anendeavour: Not only is it complicated to gather and clean a dataset of this scope, methods withwhich large-scale text analysis is possible have only recently become available. I have gathered all articles depicting domestic protests from eight major UK national newspapers published from 1992-2017 (n = 73, 654). The choice for newspaper data as well as thetime-frame was made for the same reason: in a rapidly changing media landscape, newspapershave prevailed so far, despite the internet and social media becoming mass phenomena sincethe early 1990s. Newspapers therefore allow for a long term analysis in a rapidly changingenvironment. To assess the patterns in the reporting, I use an innovative approach to identify and measureframing: I combine a manual coding method suggested by Matthes and Kohring (2008) withtext-as-data methods to be able to process coding on a larger scale. To identify frame elementsI use available literature as well as the data itself. After frame elements are conceptualised andmanually coded, the frames are revealed through cluster analysis. Machine learning algorithmsare then trained on the coded material and replicate the coding decisions on the larger dataset. First results show that the discussion is determined by three competing frames: ”Spectacle”, ”Clash”, and ”Struggle for a (just) Cause”. REFERENCES Matthes, J. and Kohring, M. (2008). The content analysis of media frames: Toward improving reliability and validity. Journal of Communication, 58(2):258–279.