Emotions on the stock exchange. Can investors’ sentiment be measured? Yes it can! It currently is done on the basis of surveys (e.g. AAII Bull and Bear), a set of complex measures (Fear and Greed index) or on an analysis of buying trends (Investor Sentiment Index). The latter means predicting the actions of stock buyers on the basis of… these actions, not an external indicator. Is it possible then, to measure emotions on the stock market? What does that mean? You will find out the answers from the below article.
Research to date on the use of sentiment in forecasting stock market indices
In December 2018, the publication “Media Sentiment and International Asset Prices” appeared, as published by the National Bureau of Economic Research. In the read the authors analyze Reuters’ and Bloomberg articles in terms of their impact on major stock exchange indices. Researchers discovered interesting bonds between sentiment in the articles and index values, which shows the example below:
- The first mood of the news correlates 4 times more strongly with the bear market (dominated by declines), than during the bull market boom,
- The second mood index is a much better reflection of the behaviour of world indices than the commonly used VIX (CBOE Volatility Index),
- and the positive mood in the media has a more positive impact on the developed markets.(but negative on the emerging markets).
Unfortunately, none of the approaches allowed for forecasts were more than 55-60% accurate. This is probably not possible while based on sentiment or emotional tags offered by social networking platforms like Twitter or Facebook.
Emotional monitoring for stock price forecasting
What if, whilst using deep neural networks, we combine media and emotion monitoring? With Sentimenti tools we have many more possibilities – up to 11 emotional variables. Our tools allow automatic analysis of sentiment and emotion in texts. We check what the sentiment level is and share each of Plutchik’s 8 basic emotions in what Internet users write during a given period and about a particular listed company, its products, services, etc. We measure the arousal of each statement.
We see if the readers like it, show their anger or loathing, and talk about the brand trust. We can even measure how their emotions change from day to day, week to week, one settlement period to another.
Then we compare the results of the emotion analysis with some hard indicators: WIG-Banks index, company share prices and so on.
Emotions on the stock exchange. Forecasting on the Warsaw Stock Exchange
At the turn of November and December 2018 we tested our assumptions in practice. For a start, we downloaded all the opinion-forming text data available on the Polish Internet from about 50 randomly selected companies. We then downloaded the data for the same period and the same entities as the exchange rate data. Our team containing financial analysts and artificial intelligence specialists prepared teaching scenarios for test models. Then we finally checked their effectiveness. It turned out that in 87.1% of cases, changes in the intensity of emotions in texts about the company allow us to predict course changes! How grand is that?!
To learn more about this topic, we offer you a complete report. It shows how often we were able to predict changes in the share price. We relied only on the basis of an analysis of sentiment and emotion. Read it and check out how many cases we have been successful in. And as for a teaser, we only will mention some of important or key Polish companies, such as CD Projekt, KGHM Polska Miedź, Gino Rossi, Enea and Wawel.