Interview: Barry Hurewitz - UBS Evidence Lab
What has it been like for your firm and your clients during the pandemic?
When it first became clear that COVID-19 was turning into a global pandemic, we completely shifted gears and made gathering real-time insights on the impact of the virus our number one priority. Last year, Evidence Lab data powered over 3,000 research reports covering every sector and all geographies. The analytic capabilities and scale of Evidence Lab are really extraordinary. Our team had already been focused on the virus in Asia, so the team had a head start. We also have thousands of datasets and frameworks to draw from, but it’s really important not to begin with the data you have.
The starting point is determining what the most important questions are now, in the near-term, and over the long-term. Then, you have to think about which of those questions will be hardest for the market to calibrate and how we can uniquely advance understanding. The goal is not to do interesting work. It is to create relevant insights that advance understanding on the uncertainties. We have the advantage of working with a number one ranked research organization to help us hone in on the most important questions in a very layered and nuanced way. I can’t overstate how valuable that is in building out our data analysis frameworks.
What are the most valuable datasets for clients at this time?
I get asked this a lot. The most valuable dataset is the one that gives you a better answer to the question you are trying to answer. It isn't about datasets for us. We have thousands and thousands of robots harvesting enormous amounts of data from the internet, and that is only about two-thirds of what we have. We buy a lot of data and run hundreds of quantitative market research surveys per year. Everything is valuable to at least one question we have or we wouldn’t go through the effort and expense to incorporate it into our analysis.
There is a lot of discussion of how COVID-19 is likely to shift employees from working in offices to working remotely. However, there is likely to be an equally important shift in the nature of how investors think about and use alternative data going forward. When alternative data first started to be used, investors often would try to build models to predict company revenue for the quarter with sources like credit card transaction data or email receipts data.
Now, with COVID-19, those models—along with many of the beliefs investors have about economic behavior and markets—need to be reconsidered and recalibrated. People are using alternative data to establish fresh baselines of what the new normal is, and past patterns may not be that helpful today. If your beliefs and assumptions about the pace of recovery do not accurately represent reality, you're going to have a hard time in this market.
I saw one highly respected hedge fund shut down at the market bottom believing that we are about to have economic Armageddon. Less than two months later, the markets look dramatically different. My point is not about whether markets have it right or wrong, but that you need an accurate read on reality or your judgement may be very off—and being undisciplined can be quite expensive.
What do you mean by being disciplined?
Results are often a function of making good decisions and luck. How many of us know investors who are brilliant, but were just unlucky with a surprising event that couldn’t be anticipated? Luck can be thought of in terms of uncertainty and risk. Modern finance has provided investors with many tools to manage uncertainty and risk. However, there is always some degree of luck that cannot be managed or diversified away, no matter how sophisticated the process.
An investor’s “decision quality” can be broken down into their “decision process” and “state of knowledge.” Good decision processes allow investors to reliably assess the likelihood that a belief is going to be true. It has to incorporate not only the information that the investor knows now and their confidence in that information, but also uncover hidden information. What other information that the investor doesn’t know now that could be known can help improve their odds of being right.
Disciplined processes are good at calibrating uncertainty well and helping an investor to see the world more accurately and objectively than the other market participants. Higher quality decision processes alter the investor’s beliefs to fit the new information rather than alter the investor’s interpretation of that new information to fit their beliefs. It may sound obvious but you would be surprised by how commonly the latter is true.
I'm sure you're getting a lot of questions about "what's next," what is the "new normal." What are you looking at to answer this?
That’s a big question that needs to be broken down into more defined questions. There are questions about de-urbanization and the reordering of global supply chains. And others about shifts in consumers’ habits or what social distancing might look like coming out of the crisis. We are gathering the most important debates and turning them into central research questions that we can test.
There are many expert views out there on each of these controversies, but I prefer to think of them as hypotheses and counterfactuals. We think it is important to gather as much evidence as possible on the counterfactual, the view that is least likely priced into the market. This forces us to search for evidence in a more intellectually honest way than just trying to prove the market hypothesis. For each of these questions, we are looking for as many different sources of evidence as we can to swarm the debates. If you want to know what’s next, you can get a front row seat by watching data as it comes in. Anyone who thinks they can predict the new normal is just speculating. I think it is far safer to calibrate beliefs more quickly and accurately by having superior sources of insights and disciplined analytical frameworks to incorporate those insights into your thinking and belief system.
Can you give us a better sense of some of the work you are doing on understanding the “new normal”?
We've done several important studies on the economics of social distancing. For air travel, we created real-world simulations where we took CAD drawings of different airplanes and overlaid them with historical passenger booking data with different assumptions about social distancing. Simulating with historical data lets us more accurately determine how often passengers are traveling together or alone.
The combination of accurate assumptions about travels and actual plane layouts lets us understand in a more precise way how far away from break-even load factors airlines are. We then looked at survey data that showed that customers aren't yet planning to travel right now, as well as pricing data that showed forward curves in bookings. These forward pricing curves tell us a lot about each airline's beliefs about the future and their strategies. We brought all of that together to help answer what airline profitability might look like.
We did something similar for Disney's Magic Kingdom. Using our established geospatial techniques, we digitized the entire park and built simulations of how many visitors could be accommodated under different social distancing scenarios. But you need to be thoughtful about it—how many guests come as families vs. couples vs. individuals? How much space does a child take up vs. an adult? We used actual shoulder-width measurements and vectorized park maps to get as accurate a simulation as we could. By combining this with tracking surveys on consumer intent to travel to theme parks, we can help clients assess how full Disney theme parks are likely to be now and over time.
To assess global supply chains, we are currently tracking 17,000 ships that make over 125,000 voyages a year to follow the movement of goods. We take into account the curvature of the earth, the fact that ships don't move in straight lines, how full ships are when they are moving in and out of ports, and the types of products transported into and out of specific docks. We also look at pollution data, but adjust it for weather to tease out pollution related to heating homes, so we can see to what degree industrial centers are back on line. We can pair the shipping data and pollution data to triangulate which industries are returning back to normal and at what pace.
We have hundreds of projects like these under way related to COVID-19 to help our clients calibrate their beliefs, but this should give you a sense of the type of work we do.
Very interesting. How long does it take to get studies like these up and running?
This is what I think works to our advantage. We've been doing this for six and half years with a team of hundreds of people around the world, so we not only have a lot of data and insight frameworks organized, but we have a lot of resources that we can redeploy quickly. It is so much faster than trying to start from scratch. That speed to market and the quality of insights we can bring to bear has really resonated with our clients at this time. Size, scale, and experience really matters at a time like this.