Investing Is One Big Data Problem

Angelo Calvello, PhD

September 15, 2015

"Shouldn’t we think about changing the way we invest?” Greg Williamson 

Your summer vacation is over and now it’s time to again don your suit of armor and prepare for your Q3 investment committee meeting. Walking into the meeting, you feel the weight of the world on your shoulders. More likely than not, you have to deliver the same disappointing news and wrestle with the same challenging issues—the performance of the aggregate portfolio continues to fall short of expectations (whether measured against your benchmark or, more arbitrarily, your peers) and some (many?) of your investment managers continue to underperform their investment targets (while they still earn a portion of their fees).

This adds up to a troubling conclusion: you’re struggling to meet your investment objective, which, in turn, is placing a strain on your sponsoring organization.

The fundamental question is: why do you struggle so mightily to achieve your investment objective(s)? From my perspective, the cause is the self-limiting nature of the parameters of how we, as an industry, invest. Asset owners and asset managers alike make investment decisions using the same information sources--some form of price and economic/financial data—and the same quantitative methods (in various iterations) to try to extract value from these data.

This is like using only light crude and traditional extraction techniques to solve all of our energy problems.

It’s time for us to admit that this well has gone dry; we’ve extracted just about all the value we can from these data using existing tools. With these parameters, of course you’ll struggle to achieve your investment objectives.

We need to move beyond these parameters and explore new, complementary information sources and new methodologies for extracting actionable insights from these sources. It is time we earnestly explore whether and how what others refer to as “big data” might help us better solve our investment problems.

Cutting through all the hype, “big data” describes both the large-scale, non-traditional datasets themselves (e.g., social, news, weather, and geospatial) and the computational/analytical tools (e.g., enhanced machine learning and related algorithms) used to discover patterns and other actionable information that were previously obscured by the scale, velocity, variety, and complexity of these data.

Let me be clear: The real value of big data in investing rests not in managers mining Twitter to pick stocks but in CIOs employing the informational edge provided by big data to make better asset, style, and even manager allocations and risk management decisions (e.g., responding to regime shifts, bubbles, and ex-ante event detection).  Even a slight improvement in such decisions would significantly improve performance and make those committee meetings less stressful.

Certainly, there are substantial obstacles to accessing the potential insights, most notably cost, finding and recruiting the required talent (e.g., data architects, data scientists, econometricians), and realigning one’s culture and business objectives to ensure the optimal application of big data.

Other information-based industries and disciplines (with much different financial incentives) have overcome these barriers and successfully use big data to tackle immensely complex problems.  UPS spends over $1b annually on predictive analytics to optimize their routings, improve customer service, and predict vehicle maintenance.  The University of California’s Chief Enterprise Risk Officer uses big data to better manage business risk, deploy resources, and lower costs across its system, all of which has reduced its borrowing costs. Wal-Mart created the world’s second largest in-memory platform to gain a competitive advantage in internal merchandising and customer segmentation. 

These are not cherry-picked examples. A C-suite survey by The Economist Intelligence Unit found 48% of respondents agreed that “big data will be a useful tool among many other emerging technologies;” 23% agreed that “big data will revolutionize the way businesses are managed;” and 21% agreed that “big data is essential to deal with rapidly increasing volumes of information.”

These responses are incongruent with our industry. Few asset owners are even thinking about big data and only a handful of stealthy asset managers are using big data to enhance existing strategies.

I’m not rashly claiming big data is the panacea to our investment troubles, but rather extending the orthodoxical view that investment success is commensurate with one’s informational edge. Efficacious investors—asset owners and asset managers alike--explore the value of all available data sets and employ all available technologies to uncover investable insights embedded in these data. Isn’t the much sought-after unicorn in our industry—alpha—the result of having superior information and superior processes?

Given our persistent struggles to achieve our investment objectives and our universally held view that investing is one big data problem, then it would seem natural to integrate large-scale, complex, non-traditional datasets into our investment framework, in spite of the barriers. An endowment CIO recognized this and, in a normative tone, added that he considered it  his fiduciary duty to explore whether and how big data might aid in his job and how it just might make those committee meetings something he looks forward to.

This essay first appeared in CIO Magazine, September 2015.

Posted on September 15, 2015 .