Abstract
Thanks to their name recognition and popularity, celebrities play an important role in American politics. Celebrity endorsements could add to the momentum of a politician's campaign and win the candidate extensive media coverage. There is one caveat though: the political preference of celebrity followers might differ from that of the celebrity. In this paper we explore that possibility. By carefully studying six prominent endorsements to the leading presidential candidates in the 2016 U.S. presidential election and statistically modeling Twitter "follow" behavior, we show (1) followers of all the celebrities with the exception of Lady Gaga are more likely to follow a large number of candidates and (2) the opinion of celebrity followers could systematically differ from that of the celebrity. Our methodology can be generalized to the study of such events as NBA players' refusing to visit the White House and pop singers' meeting with Dalai Lama.
Introduction
Political campaigns invest heavily in celebrity endorsements, operating under the assumption that a celebrity’s "brand" can transfer to a candidate. These endorsements are often viewed as "information shortcuts" for voters. However, the "revealed preferences" of a fan—who they actually choose to follow on social media—might not match the public stance of the celebrity. This research by the University of Rochester addresses the gap between celebrity influence and follower behavior, specifically looking at whether following a celebrity makes a user more likely to follow a specific candidate, and whether fans of these stars are more politically engaged or more polarized.
Methodology
The researchers employed a multi-stage statistical approach to analyze the digital "Twitter-sphere" of the 2016 U.S. Presidential Election:
- Negative Binomial Regression: Used to model the number of different presidential candidates a user follows, determining if celebrity followers are more "politically curious" than the general population.
- Multinomial Logistic Regression: Categorized users into three groups: those following only Democrats, those following only Republicans, and "Independent" followers (following both).
- 3-Channel Gender Classification: To control for gender effects (like the "gender affinity effect"), the authors built a robust classifier using a list of 800 gender-frequent first names, a Convolutional Neural Network (CNN) trained on profile images (90.18% accuracy), and keyword extraction from user bios.
Datasets
The study utilizes a massive dataset from April 2016, comprising 15.5 million unique Twitter IDs following at least one of the 16 major presidential candidates. A sample set of 1 million individuals was used for deep attribute extraction. Central to the study was the follower data for six key endorsers representing different political alignments.
The selected celebrities included:
- Pro-Clinton: LeBron James and Lady Gaga
- Pro-Trump: Willie Robertson and Toby Keith
- Pro-Sanders: Cornel West and Susan Sarandon
Key Insights and Results
The analysis uncovered a surprising "misalignment" between some stars and their social media audiences.
The "Lady Gaga Exception" and Political Omnivores
Most celebrity followers are "political omnivores," meaning they follow a higher-than-average number of candidates. Lady Gaga’s followers were the sole exception, tending to follow fewer candidates than the baseline.
Deep Polarization and Echo Chambers
The Twitter-sphere was highly siloed; 92% of the 15.5 million users followed candidates from only one party. The correlation between following Trump and following Clinton was -0.51, a strong indicator of "echo chamber" behavior. Trump and Clinton had the most "committed" followers (72% and 75% followed them exclusively), whereas only 9% of Jeb Bush’s followers followed him alone.
The Clinton Misalignment
Followers of pro-Clinton celebrities were actually less likely to share their idols' political leanings than the average user. Followers of LeBron James and Lady Gaga were less likely to be "Democrat-only" followers compared to non-followers. In Gaga's case, her followers were statistically more likely to follow only Republican candidates than the average user.
Primary vs. General Utility
While celebrity fans might not share the star's general party affiliation, endorsements were highly effective within a party. For example, among Democratic followers, fans of LeBron James were significantly more likely to follow Hillary Clinton over Bernie Sanders. Conversely, following Willie Robertson (of Duck Dynasty) increased the probability of being a "Republican-only" follower, aligning perfectly with his endorsement of Trump.
Conclusion
While celebrity endorsements are powerful tools for media coverage and internal party momentum, they do not guarantee follower alignment. This study highlights that a "follow" is not a "vote," and users may follow candidates to monitor statements rather than show support. Future research could explore the causality of these endorsements: whether they convert fans or if fans simply gravitate toward celebrities who mirror their existing views. The methodology developed here provides a framework for studying other cultural and political flashpoints in the digital age.
Summary updated