Ideology Estimation, Media Slant, and Opinion Segregation: Facebook as a Social Barometer

Master Thesis, Department of Economics, National Taiwan University

We present a general framework to place different political actors including politicians, news outlets, interest groups, and the mass public all on the same ideological spectrum, using only de-identified, publicly available Facebook data. By specifying a potential ideological universe of fan pages and selecting informative users, we are able to give some new evidence and reproduce conventional measures regarding political ideal points and media slants, and also replicate ideology distribution of citizens both at national and at state levels. Unlike previous works, our procedure does not constrain to a specific aspect of political life, can generate a reasonably smooth mass ideology distribution, is time-variant, and is also topic-decomposable. This makes it extensible and useful for future research. Several new avenues of research made possible by our estimates such as election forecasting and measuring opinion segregation on social media are also discussed.

Spotlights
  • Place distinct political actors (politician, media, interest groups) on same ideology scale.
  • Generate smooth mass ideology distribution at not only national but also state level.
  • Forecast 2016 election well and can measure opinion segregation across time and issue.
  • Contributions: Add time and text dimension, smooth mass distribution, needs only open data.