Candidate Television Emotions Bulk Dataset

For those of you wishing to integrate all of the emotional data from the 2016 Campaign Television Emotions Tracker into polling, forecasting and other models and want to work with the complete dataset, we've boxed up the entire dataset and made it available for download. The 700MB GZIP file will uncompress to just over 4GB and contains one ASCII text file per emotion. Each emotion file has a very simple format: each row contains the measurement of that emotion for a given day/candidate/station triple. The first column lists the day in YYYYMMDD format, the second is the candidate name, third is the station, fourth is the number of words recorded in proximity to mentions of that candidate on that station on that day, while the fifth column contains the score. For emotions beginning with "v" the score can be used as-is, while for emotions beginning with "c" you should divide the score column by the word count column and multiple by 100 to calculate a density score. You can look up each emotion and find its full citation in the GCAM Codebook.

This is a one-time static snapshot of all of the emotional dimensions assessed from January 1, 2015 to February 4, 2016. For the latest updates of a specific dimension, use the "&output=csv" option of the Campaign Television Emotions Tracker.


Download the Campaign Emotions Dataset Through 2/5/2016.

GCAM Codebook.