Facebook Ads as a Predictor of Election Outcomes

Live Poster Session: Session Time: 2:45pm-3:45pm

Zoom Link: https://wesleyan.zoom.us/j/99352738945?pwd=QVlOeWR1YkVMWmRBK0VHSzB1N2Y2QT09

Tanvi Modi
Tanvi Modi

Tanvi Modi is a rising senior (’22) from Lucknow, India. At Wesleyan, she’s a double major in Economics and Computer Science and is interested in finance and data analysis. Outside of classes, she is involved in the Dean’s Peer Tutoring program and also tutors elementary school students online. In her free time, she enjoys traveling to new places, going on walks with her dogs, painting, and baking desserts. After Wesleyan, Tanvi plans to attend graduate school to pursue finance.

Abstract: This project tests the hypothesis that winners of political elections in the US can be predicted using Facebook (FB) ad impressions and their prices. FB ads are priced on the basis of real-time online bid auctions and the more relevant ads can reach a user at a lower price. We posit that controlled comparisons of ad prices can serve as an indicator of a politician’s popularity because the relevance of an ad is a reflection of the user’s political preferences, known to FB through machine learning models of user behavior.
Our dataset covers the Democratic primaries for the New York City Mayoral elections that took place on June 22nd, 2021. We explore factors that can influence the cost of impressions, such as dependency of the marginal cost of impressions on the previously accumulated number of impressions.  Our results show that Eric Adams – the winner of the primary before and after the ranked-choice votes were counted, – indeed was charged less for ad impressions than his opponents, meaning that he was more popular/relevant among FB users than his opponents.

Video:

https://www.youtube.com/watch?url
Your Name (Your Department)
presentation_poster.ppt-1