UBS showcased how it is using alternative data to support its research arm by asking a big question – is globalization being structurally changed? – and deploying data sets to turn this into practical outcomes.
Jeremy Brunelli, New York-based managing director at UBS Evidence Lab, explained how the firm uses multiple data sets to help analysts and investors shape a view.
In some cases, the firm can even make predictions that can guide research, but in most cases the data is more useful to question or confirm an investment idea. This can allow people to make a call with greater confidence or help them avoid a mistake.
“We like to swarm a debate with many data sets to prove or disprove a hypothesis,” Brunelli said.
Big questions, big data
He ran through the firm’s current work trying to make sense of where globalization trends are headed to show how UBS is deploying alternative data.
Globalization is a fifty-year megatrend but it has been challenged by U.S.-China frictions, the rise of political populism, and COVID-19. There are recovery signals, but are these evidence that the world is affirming the old pattern, or entering a “new normal” with different drivers and outcomes?
Fundamentally, what Evidence Lab does is not a massive departure. Analysts have always used data to get a view on what’s happening and what that means for portfolios.
The difference is that in the past, most of this data was financial statements, making much of it dependent on company reports produced at a regular but slow pace.
Alternative data is now being deployed at scale to create faster, more nuanced views, and even to predict trends. Firms that can deploy such data at scale and integrate it deftly into their research enjoy an advantage both in speed and in confidence around their investment ideas.
UBS Evidence Lab is the firm’s department preparing datasets that can be immediately plugged into analysts’ work.
How ‘alt’ is the data?
For the globalization theme, measurements of container ship activity suggest activity is picking up.
Some of what counts as “alternative” data is not very whiz-bang. UBS relies on surveys, for example: it’s useful proprietary information, but not tech-driven.
These results are augmented by genuine alt-data, such as a monitor of U.S. companies reshoring manufacturing. This too is still fairly low-tech, involving a lot of screen scraping. It’s a question of doing so across a spectrum of sources covering multiple industries.
Predictions…are the holy grail of alternative data
Jeremy Brunelli, UBS Evidence Lab
The analytics get more sophisticated when UBS builds proprietary models relying more on real-time information. For example, it has built a monitor of container ships, using data from maritime-traffic monitoring sites.
This data covers over 20,000 ships that it tracks by location, draft (how low it sits in the water, an indicator of cargo volumes), ship size and type, cargo manifests, and ports of call.
Noting that 70 percent of world trade is conducted by sea, Brunelli says this UBS algo, called Modified Deadweight Tonnage, tracks oceanic traffic across the Pacific, to give analysts a clear view of activity. Trade wars and stages of COVID are reflected in the results.
Data tapestry
Nonetheless, the data by itself doesn’t answer big questions, such as whether a recent uptick in activity represents pent-up demand from mid-2020 – and is therefore a one-off – or if it suggests a return to older trade patterns.
“To determine the big picture, we rely on a broad tapestry of data sets,” Brunelli said. For example, UBS Evidence Lab will combine its tonnage measurements with “movement” measurements.
For example, American foot traffic (are people going to the office?), where are consumers spending money, and where private jets landing in Macau originate from.
UBS combines these with other forms of data, such as app downloads, which can describe behavior like working from home, and airline prices, to get real-time looks at consumer confidence.
Can data help you make predictions?
Alternative data is not, however, a crystal ball. It doesn’t provide answers. Brunelli couldn’t actually answer the question of whether globalization is truly recovering. Rather, the data goes to analysts that might have a more specific question, like, should we buy construction material stocks, and be used to bolster their arguments.
That said, Brunelli says there are emerging cases of predictive analytics. “We can make predictions when we have a proxy to measure against. That’s the holy grail of alternative data.”
For example, the tonnage algo can be paired with other indicators used by UBS’s research teams, which is used to create predictors of Chinese exports volumes. This becomes more useful when it integrates with high-frequency types of data, matching these to slower vessel information to create a view on the economy.
But other aspects of the globalization story are not suited to predictive analytics. For example, Brunelli says the firm’s surveys of executives about their reshoring activities to the U.S. does not correlate to proxy data. “It’s too thematic and too structural,” he said.
In other words, these are anecdotal views, not big-data sets. Large firms like UBS are using alternative data in new ways, but the holy grail remains yet to be found.
