Social media increasing quants data points

Related Content Related Video White Papers Related Articles

Developments surrounding the internet mean that news is all around, but finding quantifiable data points from the general ‘noise’ is possibly becoming increasingly challenging to selectively identify as the volume of news accelerates to include non-traditional sources such as blogs and social media.

That is the gist of the argument behind software and services developers such as RavenPack, which seek to enable investors to identify market moving information, which could affect the prices of assets they hold in their portfolios.

Started as a business about a decade ago, RavenPack recently launched the fourth version of its News Analytics software, which looks at some 19,000 online news and social media sources in real time to identify events that could cause market movements.

The algorithms are designed to seek out some 34,000 listed companies, but to do so in a contextual manner.

As Armando Gonzalez, CEO of RavenPack, explains, this is based not on seeking to spot keywords, but by applying a proprietary named entity recognition. This means, for example, that if the system spots use of the word ‘Total’ in the context of an oil spill, it is the company rather than the word ‘total’ that is identified.

Another example would be comparing oranges and apples as fruit versus mention of ‘Apple’ amid discussion of a company.

Once identified, the system looks at the sentiment around the use of these words, to see if it gives a steer on the potential impact on users’ exposure to these particular companies. According to Gonzalez, the latest version of the system has doubled the number of market moving events that it can help investors detect.

For those running strategies that rely more heavily on spotting market moving factors, this type of solution is being sought out: according to Gonzalez there are a number of the better performing hedge funds that are using RavenPack currently to drive alpha as they seek to input the data as another signal in their models. Clients also include banks and asset managers.

What is important to note, however, is that the data generated is precisely that: “We are not selling buy or sell signals, but selling structured data,” Gonzalez says.

“Customers construct their own signals and relationships, and hypothesise the news will affect their assets in different ways.”

Social media

A key change seen by RavenPack in recent years has been the rise of microblogging through platforms such as Twitter, and social media more broadly.

The past year has seen a significant rise in the volume of ‘noise’ from sources such as Twitter, increasing the challenge for investors trying to find relevant data.

“Customers continue to find that professional media break news more than Twitter, because there are people specialising in breaking news on specific stocks and industries, versus Twitter becoming a bit of a repeater. How many times do people retweet an article or link based on what they have found on Twitter?”

Hence, it still seems that more value comes out of professional blogs or more traditional news sources, Gonzalez says.

“In social media the problems haven’t become easier. It becomes harder as more users tweet and more bots or automated accounts are being formed. So it’s a bigger problem as we evolve in social media.”

Quants alone are of course not enough. The way news works means that there is an inherent bias towards large cap stocks, as there will be more news about companies such as Apple or Nike globally.

However, where there is news about SMEs it tends to have the strongest impact, Gonzalez suggests. Markets are aware of flows of information around big cap companies, so news flow may have little impact. But when a smaller company, which tends not to generate much news flow says its CEO is leaving, it is the type of news that can significantly impact an equity.

Gonzalez says that RavenPack’s own research suggests that out of an index such as the Russell 3,000,  the impact of news flow is greater for the Russell 2,000 than the Russell 1,0000. However, it is also important to look at factors such as actual volume of news flow versus expected volume: if the norm is for, say, 100,000 articles globally per day for a company such as Apple, and this falls to a volume of 20,000 for a prolonged period, it would suggest Applie is falling out of the media spotlight. That in itself is data that clients may find interesting. Prolonged information sets can point to something happening to a company.

 

preloader
Close Window
View the Magazine





You need to fill all required fields!