Crisis management on the agenda? A big data approach to analyzing how the Norwegian national news media facilitated public response to the Covid-19 crisis.

In public health crises, people need information to help them make decisions about how to protect themselves and others from risk. Successful crisis response is thus dependent on the dissemination of efficacious information. As online news is where most people get the majority of their information, providing it during a crisis is a task for journalists and the news media. However, there are gaps in the knowledge about how the news media fills their role in communicating health information during public health crises. With this thesis, I sought to help lessen the gap of knowledge about how the news media facilitate crisis response by analyzing what they communicated and how communication changed over different stages of the Covid-19 crisis in Norway. I apply a generative machine learning approach, topic modeling, to analyze more than twenty-two thousand online news articles published by two of Norway's most prominent national newspapers, VG and Aftenposten. The model uses Bayesian statistics to categorize text based on similar words appearing together and their likelihood of appearing with other words. The method allows researchers to discover latent topics and patterns within extensive data, producing comparable results to human coders at a scale that lends itself particularly well to give detailed descriptions of news media communication efforts. Using the topic model results, I propose and test a method for operationalizing and analyzing how risk and crisis communication changes in news media content over time. I identified topics reflective of the coverage of the crisis according to their conduciveness to sensemaking and self-efficacy in the Norwegian public — building on theory on crisis and emergency risk communication (CERC). The concept of the creeping crisis provided a theoretical basis for differentiating the Covid-19 crisis from other crises. Furthermore, agenda-setting provided an additional theoretical lens to help better understand the effectiveness of this communication on behavioral change and response. The applied method was found to be fruitful in giving insight into how the selected news organizations covered the Covid-19 crisis in Norway. I identified 200 different topics covered by VG and Aftenposten during the first stages of the Coronavirus crisis. 75 of these topics focused on the crisis itself. Subsequent topics identified staple news topics such as sports teams, culture and movies, social issues, and more. 48 topics were identified as conducive to crisis management, and these were analyzed based on their prevalence over different stages of the crisis. The findings suggest that Norwegian news media disseminated information facilitating crisis response throughout the first 15 months of the pandemic, starting from the pre-crisis stage to the initial crisis stage and into the maintenance stage of the pandemic. Communication changed dramatically between stages. All in all, these topics reflected 12.5 % of all news coverage published by the two newspapers. The results indicate that this coverage largely reflected assumptions about changing communication needs during a crisis, but topics also reflect additional risk communication efforts resulting from the extended timeframe of the Covid-19 crisis