Using sentiment from news media to improve investment strategies

Dr. Svetlana Borovkova presents data that proves news and media sentiment can be used to improve investment and trading strategies in the finance industry.

Key Takeaway

In a recent talk hosted by the Thalesians in Canary Wharf entitled: “AI: Sentiment in News and Social Media for Investment and Trading,” Dr. Svetlana Borovkova presented data that proves news and media sentiment can be used to improve investment and trading  strategies.

Dr. Borovkova is an associate professor at the Vrije Universiteit in Amsterdam and head of quantitative modelling at Probability & Partners. Here are some key takeaways from her talk, as well as a look at some potential risks associated with greater use of AI in the finance industry.

Dr. Borovkova’s Key Findings

1. The news media does not simply reflect the markets; rather, it drives the markets

In her research, Dr. Borovkova measured how news sentiment impacts asset pricing, and found that there is a very clear correlation between media events and the cost of stocks and commodities.

Intuitively this makes sense. In the age of smartphones it is very hard for any of us to escape the news media, and the data shows that investors are heavily influenced by what they read online, following a bit of a “herd mentality”. Interestingly, in this case it doesn’t matter if the news is true or fake – if traders feel an asset is going to increase or decrease in value they will act on that instinct (note how Trump’s recent Tweet about Amazon had an immediate impact on its stock price, whether or not it was well founded).

The problem is that this is an erratic and illogical way to work as investors react to one or two major news stories rather than measuring sentiment overall, and this reactive approach creates a lot of unnecessary volatility and risk.

NEW YORK, NY - AUGUST 15: US President Donald Trump speaks following a meeting on infrastructure at Trump Tower, August 15, 2017 in New York City. He fielded questions from reporters about his comments on the events in Charlottesville, Virginia and white supremacists. (Photo by Drew Angerer/Getty Images)

2. If you’re making decisions based on quarterly results, you’re too late

Dr. Borovkova argues that a much more sensible methodology is to take an average of overall news sentiment and incorporate a sentiment score into financial models.

She uses mechanisms like comparing company news sentiment to sentiment surrounding the S&P overall to get a clear idea of public consensus around the value of an asset. Her research demonstrated a 3-month lag time between news sentiment and asset pricing, proving that most traders were waiting for quarterly financial results to be released before making a decision about whether to buy or sell stocks.

Quarterly financial results make up just one kind of news story, and there is a lot more out there in the media that can be proactively measured in real time in order to understand how the asset is being valued. An overall sentiment score takes into account all notable events that will impact public perception of a company – not just their quarterly financials.

Dr. Borovkova recommends that traders set up a threshold where they only sell a stock when their decision is backed up by news sentiment. She proves with her modelling that this approach decreases volatility and increases profit margins substantially.

Meltwater CEO Jorn Lyseggen similarly argues in his book Outside Insight that too many companies make decisions about all sorts of things based on information that is at least three months old, when they could be using real time external data like news sentiment to act faster and smarter than their competitors.


3. There is also a correlation between the volume of trades and the volume of news stories about a company or asset

This particular trend can also be sentiment agnostic. We saw this playing out last year when Bitcoin prices were rising in leaps and bounds as the media hype continued to increase. As CNN reported, looking at Meltwater’s data from news media, “Bitcoin prices have moved higher simply based on the sheer volume of news coverage… regardless of whether the reports or positive or negative.”

4. People pay less attention to the news media when stock prices are going up, but become very in tune with news sentiment when prices are falling

The reality is traders don’t really want third party information when stocks are doing well as they are confident they are making the right decision regardless. But as soon as prices start falling they pay more attention to what they read.


5. Social media sentiment is harder to correlate with asset prices because there is so much noise out there, but there appears to be a connection between the overall volume of social activity and asset performance

For example, if there is a lot of buzz about a topic, it tends to have an impact on the stock or asset price. This is where we still have some work to do in terms of correlating social sentiment with investment activity the same way we do for news media.

In theory the principles are exactly the same. If pricing is all about perception, then social media should lay bare public consensus even more starkly than editorial media, which is driven by only a few individuals and attempts to be objective. Meltwater recently acquired Datasift who is doing some exceptional work work this area, improving NLP algorithms to uncover some important insights using social media sentiment.

Potential risks of using only AI to inform trading

It’s important to keep in mind that while AI-driven insights and correlations found from media data can be extremely helpful to traders, this shouldn’t replace human insight entirely.

A report by risk management consultancy Parker Fitzgerald has issued a warning that AI could cause markets to crash if human intervention is completely removed from trading process. Investors need to be held accountable for their decisions, and if algorithms become too complex, we lose transparency and the ability to audit firms for their actions.

A recent article in FStech outlines these and other concerns and is worth a read. There is a clearly a role for regulators to play to ensure that algorithms are carefully supervised and markets don’t spiral out of control.

We are getting into some exciting territory here. Dr. Borovkova’s excellent research is helping to bear the flag for outside insight and the increasing importance of using of external data in decision-making. To get an idea of the insights now possible using Outside Insight, download a sample industry competitive report here.

Recent Articles