Greater computing power means greater potential for factor investing, finds Carlo Svaluto Moreolo

At a glance

• Big data can enhance factor-based investment strategies. 
• Analysis of databases of non-financial information can complement conventional factor analysis.
• Managers are starting to offer big data-enhanced strategies in the alternative risk premia space.
• Investors should not confuse big data with data mining.

Anew evolutionary step for factor investing is in sight, as asset managers embrace technological change that could make factor-based allocations yet more powerful.

The proliferation of big data could drive factor investing forward for years to come. Data-intensive strategies traditionally belong to the realm of quant hedge funds. But traditional asset managers can raise their efforts in developing the quantitative elements of factor-based strategies. This is thanks to increased computing power and easier access to large databases of non-financial information.

Factor investing was once a staple of the hedge fund world, before moving into the mainstream in recent years. Data-driven strategies could follow the same path.

In fact, many well-known asset managers are taking steps to link their factor-investing capacity with the opportunities offered by information technology. In simple terms, greater computing power can make the search for signals in financial markets deeper, broader and more efficient. But more importantly, big data offers the possibility to sift through huge amounts of real-life information looking for signs that enhance the evidence on risk factors.

A concrete example of how big data can improve on factor strategies is through a technique called natural language processing. The technique, which allows computers to ‘understand’ human language, has been studied since the 1950s but has recently become significantly more powerful. Thanks to natural language processing, investment managers can scan Twitter feeds, for instance, to look for information about firms.

Yazann Romahi, CIO for quantitative beta strategies at JP Morgan Asset Management (JPMAM), explains: “Natural language processing allows our team to analyse large numbers of news items, flagging those that may result in binary risk to a stock – for instance, a rumoured merger – in order to remove those from the universe, due to this concentrated idiosyncratic event risk allowing more effective implementation of the factor.”

A similar methodology is used to enhance the capture of the merger arbitrage risk premium, says Romahi. “Using natural language processing of news items, we flag those that are most likely to be related to merger deal activity. This is critical for the timely response to merger deal updates that is required to effectively implement a merger arbitrage factor,” he adds.

As the example suggests, the strongest link between systematic investment strategies and big data is currently being built in the domain of ‘alternative risk premia (ARP)’. The expression means different things to different people. It can refer to investment strategies that exploit ‘traditional’ factors, such as value, momentum, carry, size and quality, in a long-short format. This allows investors to get ‘pure’ exposure to a factor, minimising the exposure to a market’s beta. But ARP can also mean exposure to risk premia that differ from the more well-documented factors.  

In either case, big data can be applied to enhance factor recognition. Romahi says: “Where big data can have a big impact in factor investing is in improving the capture of the factor itself, as well as the trading of the factor. For example, in thinking about factor investing within equities, one of the most important aspects of building a factor portfolio is minimising idiosyncratic risk. Traditionally, of course, diversifying across a wide number of names was the main tool an investor had to achieve this. Big data is able to improve on that.”

Romahi says that big data could also improve risk management techniques. “Pattern recognition can also help in the construction of portfolios or monitoring of risk. Clustering and factor analysis techniques can help to identify risks in portfolios that might have gone unnoticed if risk management is driven by additive return attribution alone,” he says.

Applications such as the ones already described, despite being rare, are becoming more frequent. Chirag Patel, head of innovation and advisory for EMEA at State Street, says: “We have seen momentum in clients starting to think about using big data sources to enhance existing factor strategies.”

State Street itself provides big data products, particularly ones focusing on monitoring of media sentiment and online signs of inflation. “But there is room still to use more structured big data sources to understand better factor exposures and traditional asset exposures,” Patel adds.

Big data promises to make factor investing concepts applicable beyond public equities. State Street, in particular, has been working on helping clients better understand the factor exposures that are inherent in a private equity portfolio. Patel explains: “We aggregated a very large set of transaction-based data that we have on private-equity investments, because we service about two-thirds of the world’s private-equity assets. We found, using this novel data set, that industry factors and sector-timing skills exhibited by private equity managers appears to be an important driver of the private-equity premium.”

The research, according to Patel, could help clients build liquid private equity proxies that mimic sector exposures observed in private portfolios. “In our road map, we plan to undertake similar research for other alternative asset classes where similar big data sets lend themselves to thinking about factor exposures,” he says.

However, the adoption of big data can only enhance factor-investing techniques so long as a focus on true factors is maintained. In searching for clues on the market within huge databases, managers may find plenty of signals with a degree of statistical significance. Whether they have an actual explanation from a behavioural and economic point of view, is a different matter.

Investors should consider the dangers of data mining, which are potentially greater in big data. Aniket Das, investment strategist for index and factor-based investing at Legal & General Investment Management, says: “With improved computing power you need to be on your guard a bit more. You need to be sure that what you’re finding, in terms of signals, has some rationale behind it and is not driven by data mining.”

Das points out the fundamental difference between factor-based investing and quantitative investment strategies, which are defined to a greater extent by truly active management. Many systematic strategies based on big data tend to rely on proprietary research that is not published in peer-reviewed academic journals. As such, they could lack the rigour and focus on the true drivers behind factors.

Das says: “With greater computing power comes greater capacity for data mining, introducing lots of statistical biases into backtesting. There is a worry in academic research around what is a factor. There are hundreds of factors identified in academic research, but which ones are true factors? Today, we are able to run hundreds of thousand tests in a few seconds. Should we be using the same level of significance today as we did 30 years ago?”

JPMAM’s Romahi agrees: “Certainly, the ability to pick out patterns in large data sets is of great value to a systematic investor, but the challenge they face is knowing when to employ cutting-edge ideas and when the model risk introduced is unwarranted.”

Despite being a great consumer of data, the financial industry is only at the early stages of adopting big data among its conventional tools. Factor investing can undoubtedly be a productive area application for new, large data sets and innovative data analysis software. But the process could be slow and could hit obstacles.

JR Lowry, EMEA head of State Street Global Exchange, State Street’s data business, says: “Investment professionals are always trying to find better, more complete, more timely information. But the landscape of data sources is very fragmented right now. The industry need to figure out how to bring more non-traditional data sources to bear in a way that the financial community can get easier access to it.”