Scaling the Youtube Media Environment Using Network and Text Data

Soubhik Barari (Harvard)

Abstract: In an era where many Internet news-seekers prefer to watch rather than read their news, YouTube plays an important role in mass political communication, but remains entirely unstudied by political scientists. A key barrier is the lack of comprehensive, comparable measurements of media slant and ideal points of outlets (channels) and consumers (users). This is particularly difficult when popular political media choices are not just mainstream news outlets, but other types of news producers: late night comedy programs, political influencers, talk show hosts and opinion media. In this project, I demonstrate that existing estimation strategies incorrectly scale the ideological leanings of outlets and consumers. I propose a latent utility model that uses both text and network characteristics of YouTube videos to scale user media preference and channel-level media slant. I present initial results, diagnose methodological issues, and outline several possible applications.

View Poster in a New Tab