Everyone wants to be ahead of the market. If you can predict it, you can own it – making millions in the process.
We might not be able to predict the stock market yet, but Twitter is sure helping us along. According to a new study, reading into the patterns within Twitter chatter might be able to predict trends in stock prices.
Researchers at the University of California, Riverside teamed up with researchers at Yahoo! in Spain to see how – and whether – Twitter impacted the stock market.
This research is unique in that it doesn’t measure the correlation between Twitter and stock prices based on sentiment like most other studies in this field. Instead, sentiment is ignored in favor of the volume of tweets sent and how they’re linked to one another. In other words, it looks at Twitter more like a network of data rather than a group of emotional users.
Using this method, the research team was able to beat out other baseline strategies by between 1.4 and more than 10 percent, and the Dow Jones Industrial Average, over the course of a four month simulation.
Here’s how they did it:
The researchers first gathered the daily closing price and number of trades for 150 randomly-selected stocks from the S&P 500 during the first half of 2010.
They then created a computer program to filter in those tweets that were relevant to those companies at the time.
Their findings were nuanced: the number of trades seemed to be correlated with the number of tweets they labeled “connected components” – the number of distinct topics (such as the CEO’s actions, a new product and someone leaving the company) being discussed about that company that day – rather than the sheer volume of tweets. And they found that stock prices were slightly correlated with “connected components” as well.
During the time the simulation was running, the Dow dropped by 4.2 percent. Losses using traditional models of investment ranged from 3.8 percent to 13.1 percent – however, using Twitter data to guide simulated investments, the researchers lost only 2.4 percent, beating out all other simulations and the Dow itself.
They note that this model might work differently during an upswing in the Dow, and that the model took 30 days to begin outperforming the stock market.
Still, it’s interesting research for investors and analysts, giving them something other than just positive and negative sentiment to base their next Twitter investment strategy on.