What is the relationship between the diversity of our online information repertoires and political polarization, i.e., the intensification of opinion differences on specific topics and groups over time?
Our current online information environment provides citizens with a plethora of news and other content relevant to political opinion formation, which significantly varies in terms of journalistic quality and political extremity.
Politicians, journalists, and scientists speculate that the diversity and personalization of online information sources contribute to audience fragmentation and political polarization. This is because citizens increasingly share fewer commonalities in their news consumption and political opinions, leading to a differentiation of views.
However, research on the interplay between media exposure and dynamic polarization processes is sparse and sometimes contradictory. One reason for this is that traditional methods used by social scientists to measure media exposure, such as self-reports in surveys, inadequately capture online behavior.
Data Collection
In our four-year cooperative project with the universities of Bremen and Aalto (Finland) and the Leibniz Institute for the Social Sciences (GESIS), we are adopting a different approach. Our project combines techniques from computational social science with theories and methods from communication and political science. The empirical core is an 18-month web-tracking study of a representative sample of over 2000 German citizens from an online access panel. Participants are surveyed in four waves about demographics, political opinions, and media usage and agree to have their visited websites automatically tracked. By automated crawling and text analysis of these website contents, we can measure the diversity of sources, topics, and actors the participants are exposed to and relate this to the development of their political opinions over time.
With these longitudinal data, we systematically examine the interplay between online information usage and political opinions by tracking changes in individuals over time. At an aggregate level, we answer questions about audience fragmentation and the absolute and changing levels of issue-related and affective polarization. The results can complement scientific and political debates about echo chambers, social cohesion, and the regulation of online environments.
The four-year project was acquired within the framework of the Leibniz Competition for Cooperative Excellence by the HBI in collaboration with the cooperation partners Helena Rauxloh, Prof. Sebastian Stier, and Prof. Katrin Weller (GESIS), Prof. Cornelius Puschmann (University of Bremen), and Prof. Juhi Kulshrestha (University of Aalto).
Participating in the Study
Participants in the study will receive legally required information on data protection and the joint responsibility of the participating research institutions in accordance with Art. 26 para. 2 sentence 2 GDPR here (pdf).
More Information
The project has been funded by the Leibniz Association since 2022 as part of the competitive Leibniz Competition “Cooperative Excellence” (funding amount: €988,682.17). This program supports particularly innovative projects that require cooperative networking within and outside the Leibniz Association. This is intended to open up new fields of research and further promote cooperation in terms of the development of science locations and regions. The focus here is on high-risk research as a contribution to overcoming complex challenges.
The project work is divided into five work packages, each led by different cooperation partners. Comprehensive data collection (more than 220 million webpage visits, more than 1 million crawled and content-analyzed webpages, and more than 2,500 participants) was completed in the first half of 2024.
The project team is now in the data aggregation and analysis phase, actively working on several journal publications. By the end of the project in 2025, the results will be documented and published. The empirical tools and innovative method combination will be comprehensively documented and disseminated through training events, and the project data will be archived for secondary analysis in the GESIS data archive.
Subject and Objective of the Project
The idea of an increasing polarization of public debates is an intensively discussed topic in society. Digital media, such as social media platforms and online news offerings, are often assumed to play a significant role in opinion formation. Online information environments offer users a variety of information sources that differ significantly in terms of journalistic quality and political balance. Politicians, journalists, and scientists suspect that the diversity and personalization of online information sources contribute to audience fragmentation and political polarization. This leads to citizens sharing fewer common information sources, shrinking the common denominator of political reference points. At the same time, the state of research is inconsistent and has gaps. Existing research on the interplay between media exposure and polarization processes is limited and contradictory, partly due to inadequate measurement methods such as self-reports in surveys.
The POLTRACK cooperative project of the Leibniz Institute for Media Research | Hans-Bredow-Institut Hamburg (HBI), the Universities of Bremen and Aalto, and the Leibniz Institute for the Social Sciences GESIS takes a new approach, combining methods of computational social science with theories of communication and political science. We record citizens’ digital media usage using a web-tracking tool that tracks browser activities, providing information on websites accessed on computers and mobile devices. We then collect and analyze the text content of the visited websites to understand the topics, actors, arguments, and emotions participants encountered. At the same time, we regularly survey the participants about their political attitudes on selected topics and actors to capture potential polarization.
In this project, we conducted an innovative 18-month web-tracking study combined with four survey waves with over 2,500 citizens using a German market research panel. With the data collected, we can analyze how the lack of diversity in information intake relates to specific polarization patterns.
The longitudinal approach allows us to investigate the direction of influence between information intake and polarization and to verify exaggerated public expectations about strong media effects (e.g., echo chambers or consumption of misinformation, so-called “fake news”). Our research design allows us to detect subtle changes over time and provides a reliable empirical basis for scientists and policymakers in regulatory discussions.
Overall, the project makes a central contribution to assessing the general “health” of the German online information sphere and its impact on politics and social cohesion. Through our innovative method combination and direct observation of study participants, we contribute to the development of scientific methodology for capturing online behavior – while maintaining the highest standards of research ethics, data protection, and data management.
Our project is the only one of its kind in Germany. We are also not aware of any other international project that works with similarly extensive data in terms of the survey period (18 months), a similarly large number of study participants (more than 2,500 participants) and an automated survey on desktop and mobile devices. The size of the survey is not an end in itself. We are convinced that urgent questions about the role of digital media for democracy can only be answered through this approach.
Project Timeline and Results Overview
Work Package 1: Integrated Research Design and Data Analysis – ongoing (mainly responsible: HBI Hamburg)
At the project’s start, concepts from communication and political science were integrated into a theoretical model, and an international expert workshop was conducted. Databases of websites and social media accounts of relevant sources and actors from media, politics, and society were created as the basis for categorizing collected webpage visits and online content. Data analysis of survey data (AP2), tracking data (AP3), and content analysis data (AP4) will be conducted in the second half of 2024.
Work Package 2: Surveys – completed (mainly responsible: HBI Hamburg, University of Bremen, GESIS Cologne)
From September 2022 to December 2023, HBI, in collaboration with GESIS, designed, programmed, and conducted four online surveys covering political attitudes, demographics, and information sources. Additionally, qualitative interviews were conducted with survey and web-tracking participants in summer 2024 to validate the data.
Work Package 3: Digital Behavioral Data – completed (mainly responsible: University of Bremen)
A robust data management pipeline for storing and analyzing more than 14 million webpages and over 220 million visits to these webpages was created using the server infrastructure provided by the University of Bremen. Based on the list of relevant sources created in AP1, content is automatically crawled and stored. The tracking data collection was completed in December 2023.
Work Package 4: Content Analysis – ongoing (mainly responsible: University of Bremen, University of Aalto)
After collecting the web-tracking data and automatically crawling the content of relevant visited websites, the currently more than 1 million news articles and other accessed information offers are manually and automatically examined using innovative AI approaches for the occurrence of various topics, actor groups, arguments, and emotions.
Work Package 5: Project and Data Management – ongoing (mainly responsible: HBI Hamburg, GESIS Cologne)
This work package involves the constant coordination of other work packages and systematic identification, analysis, and control of risks. Project meetings and scientific advisory board sessions are organized. An internal data management strategy was developed, and an IRB approval was applied for, considering ethical guidelines in collaboration with legal and data protection experts. Survey items (AP2), methods for processing scraped online content (AP3), content analysis coding books (AP3), and aggregated survey responses and web-tracking results (AP1) will be published in the second half of 2024. Training and transfer events and the archiving of research data in the GESIS data archive are planned for 2025.
Joint Publications, Expert Workshops, and Grants
An international expert workshop on “Linking Surveys and Digital Trace Data: Best Practices, Tools, and Challenges” was held at the project’s start in May 2022 in Hamburg, featuring numerous scientists, including members of the POLTRACK project’s scientific advisory board.
Current project publications are in preparation and have been presented or accepted at international conferences:
- Rauxloh, H., Merten, L., Schulze, H., Moeller, J., & Lerch, I. (2023, July 18). The More News Sources, the Better for Democracy? Diversity of Information Exposure and Feelings of Self-Efficacy. Paper presented at the 9th International Conference on Computational Social Science, Copenhagen, Denmark.
- Puschmann, C., Rauxloh, H., Stier, S., Weller, K., & Kulshrestha, J. (2023, July 20). Uncovering Patterns in Political Search with Survey and Tracking Data. Paper presented at the 9th International Conference on Computational Social Science, Copenhagen, Denmark.
- Rauxloh, H., Merten, L., Möller, J., Schulze, H., & Lerch, I. (2024, June 23). The More News Sources, the Better for Democracy? Diversity of Information Exposure and Feelings of Self-Efficacy. Annual Conference of the International Communication Association, Gold Coast, Australia.
- Puschmann, C., Stier, S., Zerrer, P., & Rauxloh, H. (2024, June 24). Politicized and Paranoid? How Conspiracy Ideation Predicts Alternative News Consumption. Annual Conference of the International Communication Association, Gold Coast, Australia.
- Zerrer, P., Merten, L., Stier, S., Puschmann, C., & Mangold, F. (2024, September 24-27). Unveiling the Mobile Media Mosaic: Analyzing News Repertoires Combining Individual Mobile and Desktop Tracking Data. Accepted Presentation at the European Communication Conference der European Communication Research and Education Association (ECREA), Ljubljana, Slovenia.
- Merten, L., Rauxloh, H., Stier, S., Puschmann, C., Kulshrestha, J., & Weller, K. (2024, September 24-27). Political Polarization and Diversity in Online Information Exposure: A Longitudinal Tracking Study. Accepted Presentation at the European Communication Conference der European Communication Research and Education Association (ECREA), Ljubljana, Slovenia.
- Puschmann, C., Rauxloh, H., Merten, L., Stier, S., Kulshrestha, J., & Weller, K. (2024, September 24-27). Watching the Greens? Predictors and Contingencies of Partisan Political Information Seeking with Online Search Engines. Accepted Presentation at the European Communication Conference der European Communication Research and Education Association (ECREA), Ljubljana, Slovenia.
