Informing and Improving Sentiment Analysis by combining CSS with qualitative methods. Analysis of Differences in the Media Coverage on policies enacted by the German ‘Länder’ during the Covid-19 pandemic

The Covid-19 pandemic has introduced a predicament for governments across the world: They must simultaneously impose social-distancing measures to suppress the spread of the disease and encounter possible negative effects for the economic and psychological situation of the broader population resulting from those measures. While facing this challenge, governments and their actions taken are closely watched and evaluated by the public and the media. In Germany, the regional (‘Länder’) governments have been responsible for most political measures. We will take a closer look at the public discussion on relaxations and restrictions presented by the media. Do we find differences in the extent or timing of the media coverage, and does the media in certain Länder report more positive or negative than in others? Our contribution addresses this research question by combining sentiment analysis – a method originating in the interdisciplinary field of computational social science (CSS) – with qualitative content analysis on a corpus of 35,917 newspaper articles. Our main unit of analysis is the states (‘Länder’). We aggregate sentiments and associated terms and compare the values over time and between them. We bring the methods into dialogue with each other by creating a corpus of words that are associated with easing or tightening measures through qualitative analysis. Furthermore, statements in weeks with particularly negative or positive sentiment will be explored in-depth by separate qualitative analyses. By doing so, we link a research question from the field of Comparative Politics with a pressing topic and a novel methodological approach.