U.S. Presidential Election 2020

Dive into analysis on this upcoming election with enriched Twitter data and real-time graph visualization.

Colorful fireworks exploding against a black night sky, with one blue and one red display.

A surge in recent Twitter activity suggests an echo of 2016

Kineviz Data Scientist, Ana Areias, walks through data enrichment of Election 2020 Twitter data, performing demographic analysis with a PyTorch model trained on a public and racially balanced data set known as FairFace.

Line graph showing daily active participants on Twitter in the Left and Right camps from October 3 to October 26, 2020. The Left camp has higher activity, peaking around October 17, while the Right camp has lower activity with some fluctuations.

Methodology

We’ve taken a snapshot into the Twitter’s sphere - using about 1/60th of the Twitter data- to perform Election 2020 analysis. Starting with Twitter’ streaming API, we can pull in election-related tweets and define connections as users retweeting other users. From there, we draw our right and left camp analysis looking at connectivity through the following conditions:

  • User is connected to at least 3 other users of the same camp.

  • User is assigned to the camp it has the most connections with.

Two camps are first analyzed with users having more than 25,000 followers, minimizing the impact of bots. Dynamics observed did not change when including users with a lower number of followers.

Recent polls indicate that Joe Biden holds a sizable lead over Donald Trump. Some Democrats take this to mean the race is in the bag. They could be in for a rude awakening.

Twitter Demographics and the Election 2020

Line graph showing polling data from March to October 2020, with Biden leading at 51.6% and Trump at 43.1% in October.
A digital visualization of a network graph with two clusters of nodes, one in blue labeled 'Left' with 1205 nodes and one in red labeled 'Right' with 2671 nodes, connected by lines on a black background.