Estimating the ideology of political YouTube videos

Summary

This paper extends the tradition of latent ideology estimation — long applied to legislators, social media users, elites, and media outlets — to a new object of measurement: individual political YouTube videos. The authors propose recovering a video’s ideological position from the pattern of which subreddits share it, on the premise that politically heterogeneous Reddit communities reveal left–right structure through their differential linking behavior. Applying correspondence analysis to a subreddit-by-video sharing matrix, they produce video-level ideology scores that align with expectations from prior slant-measurement work.

Key Contributions

• A method for scaling the ideology of political YouTube videos as a latent variable, a content type previously underexplored in this literature.
• Bridges ideal point estimation, media slant measurement, and platform-based ideology inference by shifting the unit of analysis from actors to content items.
• Provides a measurement resource potentially useful for downstream work on YouTube recommendation dynamics, consumption patterns, and polarization.

Methods

The authors construct a matrix recording which political subreddits share which YouTube videos, drawn from Reddit posts that link to YouTube. They then apply correspondence analysis (an approach analogous to those used for legislator ideal points and user-level scaling) to extract a dominant latent dimension interpretable as left–right ideology. The technique inherits the logic of prior latent-space estimation methods but applies it to content rather than to actors.

Findings

• Cross-subreddit sharing patterns yield coherent ideological estimates for political YouTube videos.
• The results are consistent with the broader finding that social media sharing behavior encodes a recoverable ideological signal.
• Extending latent ideology estimation from actors to content units is feasible and substantively informative.

Connections

This work sits alongside other efforts to characterize political content and audiences on YouTube, including Munger2025-cz on YouTube’s political ecosystem and Ulloa2024-jm on platform-level measurement. Its methodological logic — leveraging cross-platform sharing signals to scale content — also connects to broader audience- and behavior-based ideology measurement approaches found in Bouchaud2026-lr and the slant/consumption literature represented by Green2025-ap and Freelon2024-sc.