Revealing the Localness of News Domains through Consumption on Social Media

• In Event: Changes in Local News and Their Political Impacts

Abstract

Local news has long been a cornerstone of democratic engagement in the United States, providing citizens with essential information about their communities and acting as a watchdog for local government. However, the landscape of local journalism is changing dramatically, marked by a notable decline in local news agencies, a trend often described as the emergence of "news deserts." This phenomenon raises significant concerns about the health of democracy and the informedness of the electorate.
In response to this challenge, our study presents an innovative approach to classifying news outlets as either local or national based on their geographical reach. Traditional methods of classification have typically relied on a limited set of news domains, hand-coded by researchers. These approaches, while valuable, are constrained by their scale and the subjective nature of human classification. To address these limitations, our research employs a data-driven method leveraging Twitter data collected from 2010 to 2023, encompassing over one million geo-located accounts. These accounts were identified by matching Twitter profiles with voter registration data, offering a robust and representative sample of news consumption patterns.
Our methodology utilizes the frequency with which these accounts share content from various news domains as a proxy for statewide news consumption. By analyzing the distribution of shared news content across different states, we can identify significant patterns in how news is consumed at the local level. We define national news domains as those with minimal state-specific variations in their distribution, and local news domains as those with pronounced state-specific concentrations.
To date, our method has evaluated approximately 3000 websites, providing a comprehensive overview of the relative presence of national versus local news outlets on Twitter. This classification not only corroborates findings from previous research but also offers a more nuanced understanding of the news landscape. Unlike traditional methods, our approach assigns continuous spectrum scores to all shared domains, allowing for a more detailed analysis of distribution patterns.
One of the unique advantages of our method is its ability to uncover complete localization patterns, revealing the specific states where local news domains are predominant. This granularity enables us to distinguish between news domains that are local to a single state and those that are multifocal, resonating across multiple states. Our approach is also versatile enough to classify emerging news websites that may lack clear editorial guidelines, such as misinformation sites, non-news platforms like campaign websites, and "pink slime" websites known for their low-quality, automated reporting.
Looking ahead, we plan to extend our classification method to other platforms like Facebook and Google search, further enriching our understanding of how local news is consumed across different digital spaces. We will also integrate our findings with at least five existing classification efforts, comparing and contrasting various definitions and examining domains that have been classified differently across studies. This comprehensive analysis will not only validate the growing news desert phenomenon but also provide valuable insights for future research and taxonomy development.
Finally, our study will delve into the dynamics of local news stories that gain national attention, examining the factors that contribute to their virality and the implications for our classification system. By understanding how local news evolves in the digital age, we can better appreciate its role in the democratic process and develop strategies to support and sustain it.

Authors

• Kaicheng Yang

• Burak Ozturan, Northeastern University

• David Lazer, Northeastern University

• Pranav Goel, Northeastern University