by Satish Srinivasan
This research project explores the potential of a lexicon-based classifier developed by the National Research Council (NRC) of Canada for determining the emotions and sentiments within tweets collected across different states during the 2020 United States presidential election. By analyzing the fear, joy, disgust, sadness, surprise, anticipation, anger, and trust expressed in the tweets, the results are used to generate overall net sentiment scores, sentiment scores on key issues, and predictions on how certain swing states will vote.
Using tweets collected from March 3 to October 2, the project produces a month-by-month net sentiment score for each candidate. The in-depth sentiment scores can also be used to see how major events impacted each candidate’s favorability; Trump’s suggestion to delay the 2020 presidential election and his attempt to block John Bolton’s book negatively impacted his net sentiment score, while the US economy shrinking and unemployment rising lowered Biden’s net sentiment score.
The project also analyzes and compares the candidates’ net sentiment scores on the key issues of the economy, foreign affairs, trade and tariffs, race relations, health care, and immigration. The results indicate Biden leads on the economy, foreign affairs, health care, race and immigration; Trump leads on trade and tariffs.
Tweets were collected from seven swing states: Arizona, Florida, Michigan, North Carolina, Pennsylvania, Texas, and Wisconsin. Based on the sentiment scores and recent trends, the project predicts for whom the state will vote, and compares the predictions to those of mainstream pollsters.
Battleground State | Our Prediction | Prediction by Five Thirty-Eight |
---|---|---|
Pennsylvania | Likely Biden | Clearly Biden |
Florida | Likely Biden | Slightly Biden |
North Carolina | Likely Trump | Slightly Biden |
Wisconsin | Likely Trump | Clearly Biden |
Texas | Likely Biden | Slightly Trump |
Michigan | Likely Trump | Clearly Biden |
Arizona | Likely Biden | Slightly Biden |