In Focus: How Alternative Data Is Giving Funds Greater Insight

Art by Dadu Shin

Alternative data and the methods to gather that information aren’t just useful to active traders trying to seize on short-term opportunities. Long-term investors can benefit, too.

Pension funds and other asset allocators can use alternative data to improve risk management, confirm or reject investment theses, and improve transparency in other ways not possible even a few years ago.

Dutch pension manager APG Asset Management started using alternative data three years ago at a new level, said Ronald van Dijk, managing director, head of capital markets investments, for the pension manager, which has $590 billion in assets under management.

One way APG uses alternative data is to bolster its environmental, social, and governance (ESG) investments. It uses alternative data, thousands of data sources, and artificial intelligence models to glean greater insight into companies’ ESG ratings as they search for firms scoring high on the United Nations Sustainable Development Goals. “We have a good overview now on the complete universe of companies and how they score or relate to the UN goals, and that helps us to invest more in SDIs (sustainable development initiatives),” he says.

Technology will increasingly help APG automatically process, translate, and summarize both structured and unstructured information related to companies, whether it’s corporate reports, digitized analyst calls, related satellite data, social media posts, or other information, in any language. Humans can also do this, van Dijk admits, “but you can be much more efficient and effective, get more information and get more insight.”

Van Dijk wouldn’t disclose whether alternative data has led to cost savings or improved returns, but he called APG’s use of alternative data, “our secret sauce.”

APG believes in alternative data so much that it became the majority owner of data analytical firm Entis to beef up its sustainable investing side, and create direct access to state-of-the-art data science capabilities in general.

How to use alternative data

Mars Spencer, vice president of financial markets for predictive analytics company Predata, says its asset-allocator clients are primarily interested in a better understanding of geopolitical themes and macro risks affecting local markets. Predata gathers metadata in websites, such as the statistical metrics on visits and interaction with websites, to form predictive analysis. This information can be a proxy of predictive behavior and can be a leading indicator to themes and topics important to markets that may enhance risk management or add to returns.

This data can be “an additional input to either help confirm or falsify conviction. If you’re a fundamental discretionary manager, or if you’re more of a systematic quantitative manager, think of us as an additional feature within a multifactor model,” Spencer says.

Marcel Kasumovich, chief strategist at TSE Capital, says the firm uses predictive analysis with a longer-term view, to objectively confirm or deny a thesis. “We’ve been deploying alternative data to complete the mosaic,” he says.

Kasumovich says he thinks about what specifically he cares about in the investment and puts the data in context to what he wants to know. In one example, he explained, Australian housing is important to a current portfolio it constructed. TSE’s belief has been that the Australian macro economy and housing in particular is strong, but Kasumovich wanted timelier information to make sure its view on housing wasn’t clouded by a bullish bias. Using alternative data tools to comb through metadata, the information returned confirmed TSE’s original view.

“You’re getting current information that to the naked eye would almost be impossible to gather and put into historical context,” he says.

It’s also helped him find specific things to track to help evaluate a future thesis, like Argentine pension reform, which he says will become increasingly important. Predictive analysis alerted him to greater interest surrounding Argentine pension reform ahead of the International Monetary Fund meetings and presidential elections there. “This was not something I pulled from the Predata machine, but it was pushed to me. I wasn’t aware of it, but now I can track it,” he says.

Chris Natividad, chief investment officer at EquBot, which is the advisor to an artificial intelligence ETF, AI Powered International Equity ETF (AIIQ), says the fund’s investments are chosen using proprietary quantitative models developed by Equbot with IBM Watson artificial intelligence. The fund is actively managed to find mispriced stocks and exploit the timing of them. .

Natividad says EquBot builds its quantitative models using several traditional and nontraditional structured financial metrics to define companies it looks at, such as the likelihood that a management team will execute on a plan and how the company is viewed relative to its peers, then combine it with proprietary artificial intelligence metrics to create scores.

“It helps us understand what companies are exhibiting the best financial health, but also it identifies additional trading metrics that give off these patterns that show they’re poised for additional market appreciation in the future,” Natividad says.

For the AIIQ ETF, which Natividad says is an evolution of EquBot’s first fund, the model combs through structured and unstructured data for 9,000 ex-US firms. The system focuses on four specific targets in data: volume, velocity, variety, and veracity. That helps with not just security selection, but the actual trading of the underlying investment, he says.

One example of how the system worked came a few months into AIIQ’s trading debut. At the inception of the fund, the system selected a small pharmaceutical company with little analyst coverage. The stock price had been generally flat for a few months. The company was developing treatments for heart disease and was conducting clinical trials globally. The system found data points picking up on positive sentiment, indicators, and results, signs that the company would receive continued regulatory approvals, which it did.

AIIQ’s position quadrupled its gain in a few months, and then the system sold the position based on signals the price appreciation had reached its zenith.

“We were able to get into the position very early on because of the system’s ability to recognize the different patterns associated with the company’s developments. The system used pattern recognition and different signals that were indicative of a company poised for market price appreciation,” he says.

Things don’t always go as planned, however, but Natividad says, “that’s the beauty of the system.”

He said the fund was hit during the fourth quarter of 2018 because it had long positions with exposure to Asia and technology firms, the types of companies that were hit during increased saber-rattling in the trade war between the US and China. “The system continues to learn through machine learning algorithms, and previously there were no instances,” he says.

But the fund rebounded because the system took advantage of the sell-off. “The system is quite objective to new opportunities. If it sees a shift in risk, it will make reallocations,” he says.

Dane Rook, research engineer at Stanford University’s Global Projects Center, who has studied institutions’ use of alternative data, says having access to this data could help pension funds operate in new asset classes, such as in the private asset space. For example, owners could get real-time information on infrastructure investments anywhere globally by using sensor data to monitor new-build projects and not just rely on information provided from operators. “A direct ownership stake in an infrastructure project can be much more appealing because you can assume more control over the decision-making for that asset,” he says.

Data scraping can help to identify potential managers, particularly new ones without a long track record, Rook says. “There’s only so much combing you can do by hand. But with digital troves out there for sites like LinkedIn, that can be very illuminating in making pass or go decisions on things like first-time funds,” he says.

Beware of poor data quality

Octavio Marenzi, CEO of capital markets management consultancy Opimas, says maintaining good data quality is paramount when using alternative data. The sheer quantity of data gathered means some of it can be wrong. “You have to normalize the data and you have to make sure it’s accurate. These things aren’t a given at all,” he says.

Marenzi points to credit card transactions as an example, generally straightforward information used to understand consumer sentiment. If a merchant changes how it lists its name on credit card statements, that can throw off how data aggregation tracks revenue. He says that happened with women’s athletic apparel firm Lululemon, which switched merchant identification to the address of individual stores rather than headquarters.

“You’re trying to process this data and suddenly you see Lululemon’s revenues plummet to zero because you don’t see it anymore, but that is not actually what’s happening. Now you have to go back and figure out all these different addresses and sift through them. That becomes a fairly arduous task,” he says.

That’s why these tools need people who can analyze the data to see where there may be mistakes or wrong information, says Jeremie Bacon, CEO of Imagineer Technology Group. “Algorithms are smart enough to suck up data, but you’re better at determining what’s real and what’s not. That comes down to the credibility of news sources and sites,” he says.

The right approach

Plan sponsors who want to partner with an alternative data provider need to have staffers with strong statistical and macro financial backgrounds to properly use the tools and understand the information, Kasumovich says.

Pension funds that can’t integrate highly quantitative searches into their investment plans might be better off partnering with an asset manager who uses alternative data instead, Marenzi says.

APG’s van Dijk suggests that pension funds interested in adding alternative data should start small and have a vision of how they can augment their investments.

“We did a lot of experiments to understand better the power of this kind of data and techniques. It’s a journey. You start small, you may fail, and you learn again. This takes a different style of leadership. You can’t say directly in the first 100 days of your digitalization journey, ‘this is my vision, let’s start now,’” he says.

Related Stories:

Fintech Can Improve Returns, But Asset Owners Must Also Protect Data Privacy and Cybersecurity

Fintech Gains a Footing with Asset Owners