Interesting study and finding from Andrew Lo re Twitter and FOMC
The Wisdom of Twitter Crowds: Predicting Stock Market Reactions to FOMC Meetings via Twitter Feeds
This Tiger Cub Giant Is Betting On Banks And Tech Stocks In The Recovery
The first two months of the third quarter were the best months for D1 Capital Partners' public portfolio since inception, that's according to a copy of the firm's August update, which ValueWalk has been able to review. Q2 2020 hedge fund letters, conferences and more According to the update, D1's public portfolio returned 20.1% gross Read More
Pablo D. Azar is a PhD student in the Department of Economics and Laboratory for Financial Engineering, Sloan School of Management, MIT. Email: firstname.lastname@example.org
Andrew W. Lo is Charles E. and Susan T. Harris Professor and the Director of the Laboratory for Financial Engineering, Sloan School of Management, MIT. Email: email@example.com Abstract With the rise of social media, investors have a new tool to measure sentiment in real time. However, the nature of these sources of data raises serious questions about its quality. Since anyone on social media can participate in a conversation about markets—whether they are informed or not—it is possible that this data may have very little information about future asset prices. In this paper, we show that this is not the case by analyzing a recurring event that has a high impact on asset prices: Federal Open Market Committee (FOMC) meetings. We exploit a new dataset of tweets referencing the Federal Reserve and show that the content of tweets can be used to predict future returns, even after controlling for common asset pricing factors. To gauge the economic magnitude of these predictions, the authors construct a simple hypothetical trading strategy based on this data. They find that a tweet based asset-allocation strategy outperforms several benchmarks, including a strategy that buys and holds a market index as well as a comparable dynamic asset allocation strategy that does not use Twitter information.
Investor sentiment has frequently been considered an important factor in determining asset prices. Traditionally, sentiment is measured by observing analyst estimates, survey data, news stories, and technical indicators such as put/call ratios and relative strength indicators. Two drawbacks of these indicators are that they are based on a relatively sparse subset of the population of investors and, except for technical indicators, are not measured in real time. The rise of social media allows us to overcome these drawbacks and measure the sentiment of a large number of individuals in real time. These data sources give the quantitative investor a new tool with which to construct portfolios and manage risk. However, because social media data is generated by individual users and not investment professionals, the following questions arise about the quality of this data:
• Do user messages contain relevant information for asset pricing?
• Can this information be inferred from more traditional sources, or is it truly new information?
• Can social media data help predict future asset returns and shifts in volatility?
To answer these questions, we focus on a single recurring event that reveals previously unknown information to the market: Federal Open Market Committee (FOMC) meetings. Eight times a year, the FOMC meets to determine monetary policy. The decisions made by the FOMC are highly watched by all market participants, and often have a significant impact on asset prices.1
To understand how investors on social media behave around FOMC meeting dates, we create a new dataset of tweets that cite the Federal Reserve. Using natural language processing techniques, we can assign a polarity score to each Twitter message, identifying the emotion in the text. We show that this polarity score can be used to predict the returns of the CRSP Value-Weighted Index, even when limiting ourselves to articles and tweets that are published at least 24 hours before the FOMC meeting.
We use these results to construct trading strategies that bet more or less aggressively in a market index depending on Twitter sentiment. We find that portfolios using Twitter data can significantly outperform a passive buy-and-hold strategy.