The collection of data sets so large they are difficult to process by normal means has its own name: big data. Nicholas Pratt asks if new technology could generate alpha returns.
“Big data” has become a buzzword, not least in the investment management industry where some see it as a means to predict market movements.
A recent report by BNY Mellon suggests big data will lead to new approaches in all phases of financial markets – and some challenges. For example, big data could make markets so efficient that active investment managers would need to find new methods to outperform.
The nature of risk may also change, says the report, A First Perspective: The Transformation Influence of “Big Data” on the 21st Century Global Financial System. Disappointing corporate earnings or unexpected economic reports could give way to the risk of interruption in the supply of data.
“As technology makes broader and deeper decisions, financial decision-making accountability may need to move beyond the realm of financial experts to diverse teams that include data scientists,” says Jack Malvey, chief global markets strategist at BNY Mellon Investment Management.
Big data is already being used in the investment industry to pursue investment ideas and may be useful beyond simply supporting research. Phillip Maymin, assistant professor of finance and risk engineering at the Polytechnic Institute of New York University, says big data can be used for measuring investor sentiment.
“It is very hard to do that without big data, unless you simply define investor sentiment to be some index or market return. Measuring is the first step towards predicting, but obviously not the last.”
Recent research supports the idea that big data can predict investor behaviour. In March, three economists, Tobias Preis, associate professor of behavioural science and finance at Warwick Business School, Dr Helen Susannah Moat, a computational social scientist at University College London, and H Eugene Stanley, professor of physics at Boston University, published a paper, Quantifying Trading Behaviour in Financial Markets Using Google Trends.
“The first simple idea was to look at search terms on company names and the relationship between transaction of those companies’ stocks,” says Preis. “Based on these initial results, we looked to see if that same data could be used for alpha generation.”
Preis and his team constructed hypothetical portfolios and measured the frequency of 98 keywords to explore whether search terms could be indicators of market movements, using the Dow Jones Industrial Average as the benchmark. “For example, an increase in the use of the term ‘crisis’ will be a signal for action the following week,” he says. “It is a sign of investor concern.”
The portfolios, on average, outperformed the index by 326% during the sample period of 2004 to 2011.
The team has also looked at data from Wikipedia, the free internet encyclopedia, and plan to incorporate social media sites such as Twitter, LinkedIn and Facebook in their research.
“We are looking to extend the use of big data resources available online to anticipate and predict real human behaviour,” he says. “The same data is already being used to anticipate riots or social unrest or the spread of disease. We are testing the idea that it can be used to predict market movements.”
The research could help the investment industry to broaden the types of data used in decision-making, says Preis. However, experimenting with big data requires technical expertise. “You can run into computational problems because these are huge databases and it involves techniques from natural language processing and computational linguistics to understand this information in an automated way,” he says.
“If you are receiving data from LinkedIn, Facebook, Twitter or Wikipedia, you are more or less listening to the entire world.”
Maymin agrees that there are considerable challenges to using big data, but opportunities too. “It is because the walls are high that there may be some more bounty on the other side,” he says.
Despite the enthusiasm, investment managers who are neither high-frequency traders nor quantitative analysts are yet to embrace big data.
“It is clearly an interesting theme but from an alpha generation perspective, it is not something we will be using in the short-term,” says Jane Lenton, head of business services and technology at Hermes Fund Managers. “It is more suitable for high-frequency trading firms.”
Hermes has researched big data for nine months, mostly to track investor sentiment rather than generate alpha. For instance, the firm has experimented with using computers to scan press releases, company announcements and news articles. “It is all part of the research process,” says Lenton. “With Hermes’s investment philosophy, I think big data will be limited to research use rather than alpha generation.”
Even when limited to research, there is a need to improve methods for analysing big data, says Lenton. “It can be very hit-and-miss depending on the keywords you’re looking for and the sources that you are using,” he says .
“The definition of big data can also change depending on who you speak to. From an IT perspective, we look at it as unformed and unformatted data rather than market data, which is much more precise.”
Poor performance of some computer-driven or quantitative funds may also discourage fund managers from embracing big data. The Newedge CTA Trend Sub-Index, which tracks the performance of the largest quant funds, fell 3.4% in 2012 and 7.9% the year before. One of the most high profile quant funds, David Harding’s $10 billion (€8 billion) Winton Futures Fund, lost 3.5% in 2012.
Despite these statistics, there were a record 187 quant funds launched in 2012 and there has been a 40% increase in the number of quant fund start-ups in the first quarter of 2013, according to data provider Preqin. These figures suggest the use of algorithms in fund management will only increase.
Lenton agrees that big data will be more and more useful to investment managers in future, but believes it will be a supporting tool.
“Analysts will look at big data before making any investment recommendations. In terms of asset allocation or ownership, they will look at sentiment in the marketplace on particular assets. I don’t think we are at this stage at the moment but we will keep an eye on this area.”
©2013 fudnds europe