Who do I support now: a case study of twitter user's second choice candidate after their preferred candidate withdraws from the 2020 Democratic Party presidential nomination race

Abstract

This thesis explores how tweets of Twitter users and their reciprocal friends are related to the 2020 Democratic primary candidates for which these users indicated preference, after their initially preferred candidate withdrew. A new methodology was used for collecting samples and extracting relevant features while working within the limitations of free Twitter access. The sample contained users who indicated a preference for either Elizabeth Warren or Pete Buttigieg from Nov 1st, 2019 to Mar 1st, 2020 then later indicated final candidate preference for Joe Biden or Bernie Sanders from Mar 5th to Mar 13th, 2020. Evidence was found that final candidate preference for Twitter users who self-identified as supporting an initial and final candidate is significantly associated with user and friend network tweet frequencies and sentiments for tweets that involve the final candidates’ names. Evidence was also found that initial candidate preference is significantly associated with final candidate preference. Logistic regression modeling was used to explain new candidate preference. A dominance analysis was conducted and found that weighted friend network tweet frequencies was the strongest factor in predicting final candidate preference, followed by the difference in user tweet frequencies, and lastly the candidate for which the user initially indicated support.