The 45th President of the US, Donald J. Trump, often addresses a wide variety of topics concerning public companies, countries and future U.S. policy on his Twitter. From the morning of his inauguration through its one-year anniversary, Trump has sent out 2,568 tweets, which is a bit more than seven tweets per day [1].
Trump’s frequent activity on Twitter has certainly made his Twitter feed an important market-moving force. In 2019, JP Morgan Chase launched the Volfefe Index due to the statistical significance of Trump tweets on market sentiment. Their analysis confirmed direct correlations between Trump's tweets and subsequent market movements [2].
The impact of presidential tweets on the U.S. equities market has been an ongoing topic of discussion in the financial community. I discovered some systems that were built to spin that effect.
With a team of 5, I initiated the project to visualize the correlation between Trump's Twitter feeds and the stock price. In addition, we wanted to experiment with the idea of stock price forecasting from Twitter feeds.
Data Visualization
We built the following website with data from Wharton Research Data Services (WRDS) and Trump Twitter Archive. It visualizes the movement of S&P 500 Index influenced by the president's tweets from 2015 to 2018. The website is interactive that allows viewers to zoom, inspect, and link to the original tweet.
Prediction
We also built a prediction model that forecasts the stock price within 30 minutes of the tweets being posted. Below is a demo of prediction on the testing data.
With the LSTM deep learning model, we achieved 61.23% accuracy on backtesting. It performs better than any other sentiment models.
Limited Influence
It's worth pointing out that the influence of Trump's tweets on the equities market usually goes away in a couple of trading days and doesn't carry much long-term effect [3].
Credits
Binghong Chen (Machine Learning), Jiahao Cui (back-end), Yincheng Zhao (back-end), Yujia Wang (front-end), Chu Han (data-processing), Cheryl Jiang (Design)
Comments