Twitter is not just for chatting, but can also predict the financial markets? The Federal Reserve creates the Twitter Financial Sentiment Index (TFSI)

share
Twitter is not just for chatting, but can also predict the financial markets? The Federal Reserve creates the Twitter Financial Sentiment Index (TFSI)

The Federal Reserve of the United States has released a paper titled "More than Words: Twitter Chatter and Financial Market Sentiment," which examines the relationship between Twitter conversations and financial market sentiment, providing valuable insights for investors and analysts.

Twitter Financial Sentiment Index TFSI

This model uses natural language processing technology to conduct sentiment analysis on financial-related terms and phrases on Twitter, resulting in the Twitter Financial Sentiment Index (TFSI). When compared with other financial condition indicators based on prices and surveys, TFSI shows a high correlation with financial market performance.

Data is collected from 2007 to April 2023, totaling 4.4 million posts. A higher TFSI index indicates deteriorating sentiment. The article mentions that after 2017, this index provided more reliable information. As shown in the graph below, the index rises as pressure on the U.S. financial markets increases, including events such as market sell-offs related to emerging market pressures in 2014, turmoil triggered by concerns over the Chinese economy and significant oil price declines at the end of 2015 and early 2016, the U.S. economy entering a recession in August 2019, the COVID pandemic in 2020, the Russia-Ukraine war, and the Silicon Valley bank collapse.

The article also compares the bond yield spread with the TFSI index, showing a high degree of similarity.

The model further proves that the overnight TFSI index, using tweets from t-1 4 p.m. to date t 9 a.m., helps predict stock market returns. The index also aids in predicting information on changes in U.S. monetary policy and can forecast tightening actions before the FOMC statement is released.

It seems that with the increasing number of social media posts, even the content of these posts can be analyzed and predicted using language models, providing valuable insights for investors and analysts. Even the Federal Reserve has recognized the importance of social media.