Climate change is one of the biggest societal and political challenges today. The well-being of future generations depends heavily on the mitigation and adaptation policies adopted today, especially by high-emitter countries, such as Germany.

The media can influence public perception of climate change and policy-making by selecting and prioritizing issues, as well as serving as a source of information. Thus, media coverage can shape the acceptance, support, and output of climate policies, as well as personal behavior, such as mobility choices. Frequent exposure to news that highlights the personal implications of climate change, such as local extreme weather events, has been associated with increased concern about climate change. Additionally, the position of the reporting matters as it reflects the importance assigned to the issue (e.g., the first news item on a television program or the front page of a newspaper). Therefore, it is important to study how climate change is covered in the media. In our study, we focus on three dimensions of climate change coverage: frequency, position, and content. Additionally, we examine how climate change coverage compares to the coverage of other news topics.

Developing an algorithm to leverage television news in content analysis

Television remains the most widely used source of information worldwide. However, most previous research focuses on text-based news as it integrates more easily with the established methods of news content analysis, which are largely text-based. As a result, television news is harder to study and has received less attention. Therefore, we set out to develop a method that would make television news more accessible for analysis. This project started out quite exploratory as a research internship for my undergraduate degree in computer science. It then turned into my bachelor thesis supervised by my co-authors Jakob Wedemeyer, Annika Stechemesser, and Leonie Wenz. 

I started out by gathering a dataset on the German television newscast Tagesschau. Tagesschau is the biggest German television newscast with a market share of roughly 40% (in 2023). The dataset contains the subtitles of Tagesschau shows from 2015 to 2023, as well as the headlines of all the news items covered in each show. Next, we spent a long time developing, testing, and tuning an algorithm that splits the subtitles of each show into the individual news items. The idea was to extract important keywords from the headlines and try to match them to keywords in the subtitles. Important keywords from the headlines tend to appear at the beginning of the news item. This allowed us to identify the beginning of news items in a show and split the subtitles into individual news items accordingly. Using the final algorithm we matched 95% of headlines in the data, totaling over 28,000 news items with corresponding time stamps.

Classifying news topics with neural topic models

Next, we classified the content of the news items, leveraging a combination of a dictionary-based approach and neural topic modeling. Each news item mentioning a climate change-related term was classified as climate change coverage. The climate change-related terms stem from a dictionary of keywords compiled based on prior literature. While this approach allows us to target climate change news items, it is not suitable to classify all other topics discussed in Tagesschau as we would need a dictionary for each topic. Therefore, we complement the dictionary-based classification approach with neural topic modeling, a method of unsupervised machine learning that incorporates large language models. It discovers topics in the input data without any prior knowledge of the present topics, enabling us to flexibly classify all other Tagesschau news topics.

Climate change coverage is increasing

Overall, we find that climate change coverage constitutes 4% of Tagesschau's total news coverage from 2015 to 2023, with an increasing trend. This aligns with climate change coverage in German newspapers. On average, climate change is covered in an intermediate news slot. Acute crises in particular surpass climate change coverage temporarily in terms of frequency and positioning. Three major crises occurred within the time frame covered by the data: The so-called migration crisis in 2015, the COVID-19 pandemic 2020-2022, and the war in Ukraine from 2022. Each of these crises received over 30% of the total news coverage in individual months. Additionally, we find that sports are covered more frequently than climate change, but placed in later news slots.

Climate change coverage is focused on climate policy rather than climate impacts

To gain understanding of which climate change topics are covered, we examined the thematic composition of climate change coverage, i.e, all news items that include climate-change related terms. We find that approximately 80% of climate change coverage is focused on national and international climate policy. About 10% of climate change coverage reports on climate impacts. About 8% report on climate activism, such as the Fridays for Future protests. Moreover, climate policy tends to be discussed at the beginning of the program, while climate impacts and climate activism are discussed toward the end. The frequent and prominent coverage of climate policy indicates a higher news value than climate impacts and climate activism. One possible explanation for the later positioning of climate impacts in the show is that they may be grouped thematically with the weather forecast at the end of Tagesschau. This may create a more coherent story-line in the show. Nonetheless, the finding is interesting, given that previous research shows that reports on extreme weather events raise public concern about climate change. Conversely, framing climate change as an international problem may lower public engagement, as people may feel powerless. Additionally, our findings for climate impacts do not align with our findings regarding the frequency and positioning of other acute non climate-related crises, meaning that climate impacts are reported differently.

Upon closer examination of the climate impacts, we find that floods and storms are covered most frequently and most prominently relative to other climate impacts. However, reports on climate impacts that do not draw a connection to climate change by mentioning climate change-related terms were not included in this analysis. Nonetheless, this finding could be explained by the fact that Germany has been severely affected by floods and storms, as they have caused the highest reported economic damage among all climate impacts in Germany thus far. In contrast, heat rarely makes it into the first half of the program, despite being the deadliest climate impact in Germany.

Conclusion

The media—and television in particular—can influence public perception of climate change and policy-making. This study demonstrates a way to access and leverage television data for content analysis. Specifically, we employ a combination of natural language processing and machine learning methods.

Our findings show that climate change coverage has been increasing on German television over the 2015-2023 period, totaling 4% of the total new coverage on Tagesschau. Moreover, we find that climate change coverage is largely focused on climate policy, which is covered frequently and prominently. Climate impacts are covered at a low frequency and in later news slots. This contrasts with the high economic and societal damages they cause. Previous research indicates that more frequent exposure to climate change news potentially increases public concern about climate change. However, a focus on personal implication, such as local climate impacts, has been found to be particularly important to reduce the psychological distance to climate change. 

Overall, we hope that our study will make television news more accessible to future research and provide interesting insights into climate change coverage on German television. We also hope that it will allow future research on further factors that may impact public perception of climate change, such as framing and sentiment.

Photo by Oguzhan Tasimaz on Unsplash.