Parking Lots Don't Tell the Whole Story: The Trouble With Alternative Data
(Bloomberg Markets) -- It’s simple, really: To beat the market, just have insights before everyone else. Some try to stay a step ahead using brute force. Hedge funds across the globe have spent billions on super-high-speed internet connections and prime real estate to place their servers within spitting distance of an exchange. Others go for guile. Hire the biggest brains and let them (or their algorithms) slice and dice, correlate, and sample widely disseminated data to identify trends and relationships the competition can’t. Then there are those who’ve tried to go to the frontiers. Let everyone else scour the well-trodden ground of sales or earnings figures; they’ll look at alternative data such as satellite imagery of parking lots to track sales before they’re reported. Problem is, the alternative data space is looking a bit well-trodden itself. So while finding an undiscovered signal remains the grail, seeking it on the data frontier may bring its own set of difficulties.
Where do you start?
“Having a wealth of data is great, but only if you really believe it is going to improve your ability to forecast and capture market inefficiencies or risk premia. Available information is not synonymous with useful information.”
—Ray Iwanowski, Managing principal and co-founder Secor Asset Management LP
What makes data useful?
“Our work relies on data that covers at least one full economic cycle, but we prefer an even longer-term perspective whenavailable. The data that we find is most impactful for our quantitative investment team includes thousands of companies, compared against each other, on a plethora of points.”
—David Blitz, Head of quantitative research Robeco Institutional Asset Management BV
It’s not one-size-fits-all?
“It’s important to match the data set and resulting signal to a manager’s style—specifically, their breadth and timehorizon/turnover. An event-driven strategy using retail sales receipts data would typically cover a narrow universe of retailers and consumer products companies. The resulting signal would be low-breadth and likely work over time frames of days or weeks. That wouldn’t be suitable for a diversified, lower-turnover investment process. Brand sentiment data could produce a slower-moving set of signals useful for lower-turnover managers but, again, would typically be available for a limited universe of companies.”
—John Chisholm, Co-CEO, co-chief investment officer Acadian Asset Management
What’s worked for you?
“Alternative data sets have been useful for evaluating potential turning points in real time, such as monitoring credit card transaction sales at Nike following the launch of the Colin Kaepernick ad campaign. There are other potential uses for alternative data sets, such as to help build a case for why Amazon or Wayfair may continue to trade at high multiples. In contrast, traditional quant value spreads have had well-publicized challenges from shorting names like these.”
—Ray Carroll, CIO Neuberger Berman Breton Hill
Is there a rule for what to avoid?
“We stay away from over-marketed data purely curated for hedge fund consumption, such as satellite data, credit card transactions, and email receipts. These data sources are overused, and we have seen a marked deterioration in theirpredictive power.”
—George Mussalli, CIO and head of research, equity PanAgora Asset Management Inc.
Environmental, social, and governance data is one of the most widely known alternative data sets. How useful is it?
“Issues-based data is a more reliable and finite indicator compared to ESG ratings, because it typically measures something concrete, i.e., women board members or carbon emissions. This type of data allows you to judge whether or not it is a positive, negative, or neutral attribute. By contrast, broad-based ESG ratings are subjective and made with varying criteria. What makes a ‘good’ or ‘bad’ ESG company varies from analyst to analyst.”
—Zachary Wehner, Portfolio manager Crossmark Global Investments Inc.
What data hasn’t lived up to the hype?
“Fund flows are probably the mostoverrated data set when it comes to managing investment product design—not because the data is in any way meaningless, it is in fact very powerful—but because the data set is often stale and used in after-the-fact market analyses by market followers, and not market leaders. The firms that lead the marketplace instead deploy multifaceted strategies that use focus-group data, qualitative market-sentiment analyses, and statistical machine-learning techniques that seek to predict the next ‘big’ product for investors.”
—Ronan Brennan, Chief product officer Compliance Solutions Strategies
How do you evaluate a new source before implementing it?
“Availability: If the vendor is selling the data to me, doesn’t that mean many other people also have access to it? Integrity: If the vendor’s goal is to sell the data, what is their incentive to avoid overfitting or letting forward-looking information leak into the data set? Market perception: Asset prices reflect market participants’ consensus perception of fair value. If this new data set is, in fact, one which no one else has, it’s still necessary for this information to find its way into market perception and ultimately impact prices. If it doesn’t, the insights the data provides may not result in the expected price appreciation, and in fact, you might be carrying some risk along the way in trading on it.”
What’s a trait of data that you find valuable?
“We’ve been incorporating big data in our models for over a decade. Our best data comes from unknown data sources or data we collect ourselves.”
Satellite imagery has come up a lot.
“While the data is very cool, there aresignificant challenges in transforming this into actionable investment insights. The breadth of this data can be low, the historical frequency and resolution available can make back tests difficult, and the precision of some of the data in terms of predicting the variables actually of investmentinterest—for example: How consistently does foot traffic translate into retail sales? And even if it does, is the earnings impact already discounted by analysts?—is limited. For many institutional equity managers andinvestment styles, satellite imaging data still has limited investment value, though it is gaining increasing use among commodity traders.”
What’s a good first step when thinking of working with big data?
“What’s most important is to sift through all of the noise to find what data is actually relevant to the work that each investor or investment team is trying to accomplish.”
Are there characteristics that make a data set less useful?
“We’re less excited about data sets where the connection to financial metrics is indirect and where the breadth of coverage limits the ability to make relative comparisons, such as satellite imagery of mall parking lots. We think there are richer opportunities. More common data sets that have been on the radar of investors for a long time, such as insider transactions data, are still widely used, but we see less opportunity to provide an edge over what the market already knows.”
Kochkodin is a managing editor at Bloomberg News in New York.
©2018 Bloomberg L.P.