This project is part of research work at the University of Connecticut-Stamford Campus. People: Arjun N Patel, John Iacovacci and Philip G. Bradford
A company is any business whose stock is listed on a U.S. stock exchange or whose securities are publicly offered in the United States.
Business outlook is a qualitative measure of how a company believes it will perform financially, especially in the short-term. For example, if a company believes its new product will have a record number of sales, that is a positive business outlook. As another example, if a company thinks that it will experience supply-chain issues, thus taking loss in the near future, that is a negative business outlook. The term outlook is also used. The goal is to quantify this and see how it causes changes in stock value.
How can sentiment analysis of a 10K filing or earnings call (an event) predict how the stock price will change according to the event? For a given company, collect text from legal filings with the U.S. Securities and Exchange Commission and earning calls transcripts. Preprocess this text with text classification, sentence tokenization, and other methods. Run the processed text through a custom sentiment analyzer. Use the generated value as an explanatory variable for percent total return on company’s stock in a linear regression. Evaluate if sentiment is a good explanatory variable for percent total return.
- If a company is generally positive and/or optimistic about its business (positive outlook), the price of the stock will increase. If a company is generally negative and/or pessimistic about its business (negative outlook), the price of the stock will decrease. Justification: a company that thinks it will do well in the future acts more confident in front of investors. As a result, investors will drive the price of the stock up. A company that does not think it will do well will act less confident, making investors nervous or skeptical. This could drive the stock price down.
- All companies will describe their outlook truthfully in a legal filing with the SEC, such as a 10-k report. Justification: a company that lies about their financial or economic situation in a report to the SEC might face severe legal consequences. Most companies would not take that risk. So, they would provide the truth about their business outlook.
- All companies will describe their outlook truthfully in an earnings call. Justification: if a company lies to their investors or romanticizes their situation, and then does not perform as expected, investors would lose trust in the company. This could cause social and economic consequences for the company and would possibly incur a loss greater than simply admitting to the true expected earnings. Most companies would take the lesser risk and answer questions on earnings calls with the truth.
Sentiment analysis is any algorithm that takes in a piece of text and returns some quantitative data on the emotional value of that text. SEC legal filings and earnings call transcripts will provide a significant amount of text to analyze. If the sentiment analysis algorithms deems that the majority of the text has positive emotional value, the company clearly has a positive outlook. Thus, we can infer the stock price will increase. By the same logic, if the text collected has negative emotional value according to the sentiment analysis algorithm, the company has a negative outlook (the stock price will decrease).
The hypothesis for all the models: The sentiment score of a text and the percent total return, where Original Date is the date of the 10-K filing or Earnings Call, are related in a linear regression where the line has positive slope.
Above is an edited introduction to the report. To read more of the report, please see this Google Document: https://docs.google.com/document/d/1WkpPFeueW7tvD0onlND1vFPg1u333fbBOKa2XpSzYnc/edit?usp=sharing