28.03.2024

Social sentiment

Not long ago I came across an interesting article about how social media affects the value of assets. The authors of the article analysed tweets related to companies and ran them through their own neural network to estimate the connotation of the tweet itself. The article does not provide many calculations, but it does suggest that the impact of social media on the market today is very large.

The market as a whole is a very complex structure and there is no single model that accurately predicts market behaviour. There is a set of factors that traders and funds use to make decisions, but statistically speaking, the factors used to interpret the price movement of assets explain only a very small part of the movement. And this is a problem on the one hand, when on the other hand it is a natural situation. If there were instruments, explaining and predicting the market movements with a very high accuracy, there wouldn’t be the market, because everybody would know everything.

And that’s why new theories and all possible predictors, specifying the existing financial models, appear every day. Even the same robots in the market are able to work only in a narrow range of price and time — for the rest you need a human.

After reading this article, which analysed the price movements of 4 companies — Vodafone, EA, Cerner, T-mobile — I decided to try to analyse such data myself.

Unfortunately I could not get the data directly from Twitter, but I took the data from Bloomberg DB. Since 2016, Bloomberg has been integrating a stream of data from Twitter that relates to various traded companies. Bloomberg also analyzes Twitter data for positive and negative connotations, which seemed basically handy enough when collecting data. I did not want to concentrate my analysis only on some particular companies, on the other hand I wanted to find the correlation between asset’s price and tweets.

Thus, there was a question — what companies to take for analysis? After a brief googling I found something called MEME-index. Essentially it is an ETF on companies which are the most influenced by social media information, according to the creators from Roundhill Investments. There are 25 companies, including Robinhood, Rivian, Carnival, GameStop, AMC, etc. The index wasn’t launched until early 2021, just after wallstreetbets via Reddit was able to openly turn the market against several hedge funds and drive GameStop’s stock up against any analytics.

In addition to this list, there were taken companies that are supposed to be on the radar and frequent contributors to tweets — Tesla, Apple, Nvidia, Netflix, etc. Twenty-two more companies.

The result of all the analysis was that, in fact, tweets have no effect on profitability. Neither negative, nor positive, nor the total number of tweets. However, it turned out that tweets have an indirect influence on volatility — the more activity in twitter, the more volatile becomes the asset itself. This influence is rather short — it occurs during the day and does not spread further, but it is there in general.

We are only at the beginning of the way and we still have a lot to analyse in order to get more robust results, but already now by this preliminary analysis we may surely state that social networks influence not only 14-year-old girls’ depression but financial markets as well. And not just on stocks of billion-dollar companies. Social media can be used to manipulate and deliver critical information. It means now any investor should be as attentive and cautious as possible when analysing the information, which comes to him. Even an initial analysis tells us that it is statistically impossible to make money by reading tweets. But it’s easy to catch more risk in your assets — through increasing volatility of the asset about which a lot is written in twitter.