Using Facebook data to predict the 2016 U.S. presidential election.

Area of Science:

• Computational Social Science
• Political Science
• Social Media Analytics

Background:

• Understanding user political leanings is crucial for election forecasting.
• Traditional polling methods have limitations in real-time data acquisition and potential biases.

Purpose of the Study:

• To develop a low-cost, real-time election forecasting method using social media data.
• To measure dynamic ideological positions of users and political entities on Facebook.
• To predict the 2016 US presidential election outcomes at the state and national levels.

Main Methods:

• Utilized 19 billion likes from top 2000 US fan pages on Facebook (2015-2016).
• Derived dynamic ideological positions for politicians, news outlets, and users.
• Calculated state-level candidate support rates based on user ideological alignment.

Main Results:

• Facebook data predicted Trump winning the electoral college and Clinton winning the popular vote.
• State-level Facebook support rates demonstrated similar trends to state-level polls (cointegration).
• Polls showed smaller margins of error but often overestimated Clinton's support in conservative states.

Conclusions:

• Social media likes offer a viable, low-cost method for real-time election forecasting.
• Passively revealed preferences on platforms like Facebook provide valuable insights into voter behavior.
• This approach minimizes researcher discretion and offers a scalable alternative to traditional polling.