As the Covid-19 outbreak spread from February into March, it made sense that hospitals would face massive challenges around maintaining enough staff and equipment to handle an influx of patients. Still, it might’ve seemed antithetical that a global health crisis could be ruinous from a business perspective.

UBS’ Evidence Lab has created a proprietary dataset that shows that for all hospitals reporting so far—17% of US hospitals—average surgical volume was down 26% in March 2020 compared to March 2019; emergency room visits were down 16%, and outpatient and inpatient discharges were down 11%. As procedures like endoscopies, colonoscopies, and bariatric surgeries are being delayed or canceled, hospitals are delaying capital expenditures, including robotic surgery, radiology, and mammography, according to UBS.

In England, the British Heart Foundation (BHF) recently said the number of people visiting emergency rooms in the country with heart attack symptoms “have dropped from an average of around 300 per day at the beginning of March, to around 150 per day recorded by the end of March. A 50% drop is equivalent to approximately 5000 of the expected people every month, or more than 1100 people every week, with possible heart attack symptoms not being seen in emergency departments.”

Hospitals depend on day-to-day procedures, elective surgeries, and, though macabre, admitting patients to emergency rooms. A hospital collapses, and it has a far-reaching ripple effect on its local community.

Looking at the migration data, we saw a clear recovery [in China] out of the bottom in early February.

Jeremy Brunelli, UBS

For hedge funds looking to monitor the health of hospitals, geolocation data—both GPS and satellite—can be used to monitor footfall traffic in and out their doors—including both workers who are there for longer stretches of time, and patients coming in and out for tests and procedures.

“Hospital bonds are an important part of the muni environment,” Dean Barr, head of bespoke projects for alternative data provider Eagle Alpha, tells WatersTechnology. “Since elective surgery is on hold, studying traffic back to these hospitals is important for bond pricing, default risk, and recovery rates. It is also the case that the community suffers since the largest employer in many communities are the hospitals.”

A Delicate Blend

Buy-side firms are also analyzing footfall data to spot activity before any other lagging financial metrics, thus indicating a Covid-19 recovery, Barr says.

“Looking at an economy that’s just completely shut down, the first thing you’re going to want to look at is any information or service that can provide you with some sense of footfall. That footfall traffic, whether to a mall or a restaurant or a gym—to anything that has been previously shutdown—can signal some form of restart, however small or large it is,” he says.

Geolocation takes data from cars, cellphones, and computers connected to the internet, and connects that to an anonymized person. Geolocation data service providers often use a blend of satellite imagery, GPS data, Wi-Fi, and IP addresses or DNS activity coming from computers and phones connected to the internet, Barr says. Finding actionable insights in geolocation data can be difficult and there is no one “silver bullet” dataset; there’s a certain blending of potentially multiple geolocation datasets along with other forms of alternative data, such as credit card, weather, sentiment, and/or payment data.

During a webinar on April 23, Jeremy Brunelli, head of Evidence Lab frameworks for UBS, showed how the bank is combining geolocation outputs to track the coronavirus recovery in China. For example, the evidence lab team took the China Migration Network Traffic model, which primarily tracks mobile phone movements for Baidu users and shows people’s movements into and out of cities and provinces in China, and they paired that with UBS Evidence Lab’s Global Traffic Congestion Monitor, which takes in data from TomTom navigation systems to monitor traffic on a road or highway.

“Looking at the migration data, we saw a clear recovery out of the bottom in early February,” Brunelli said. “It recovered in the March timeframe. It took a bit of a step down … and then it settled down at a new plateau a bit below that original plateau.”

For the broader Asia-Pacific (APAC) region, they took the Congestion Monitor and combined it with the Global Public Transit Traffic Monitor, which takes data from the mobility app Moovit to see how people are using buses, trains and subways. Brunelli said that Hong Kong and Australia were showing signs of a bottoming several weeks ago and has since shown signs of improvement.

• READ MORE: The combination of 5G and the Internet of Things (IoT) will see data volumes skyrocket. As a result, Anthony Malakian says data providers will be able to find new datasets to package for clients, while investment firms can create unique investment insights. Click here to read more.

In the Amazon

Advan Research was started in 2015 to analyze geolocation data for hedge funds. Its CEO, Yiannis Tsiounis, says that the vendor takes in about 6 billion data points per day from about 13 million cellphones and 1,000 cellphone apps.

If you have an app on your phone that you’ve given permission to share data, the cellphone, even while resting in your pocket, periodically pings a cellphone tower. In the US, there are about 175 Amazon fulfillment centers or warehouses. Advan sees about 10,000 individual devices in those warehouses. If that device stays in the warehouse for four hours or more, they’re assumed to be an employee.

Then, similar to how a Gallup poll is conducted—where 1,000-or-so-people are surveyed, and that information is extrapolated to the entire US population—Advan takes information from those devices inside Amazon warehouses, compares it to its larger dataset of 13 million cellphones and to the broader US population to develop an estimate as to how many people are actually working inside an Amazon warehouse, as some people don’t carry their cellphone, have it turned off, or don’t share data.

“The peak [of employees working in Amazon warehouses] has been as high or higher than [the December] holiday season, he says. “So Amazon has more daily employees now in their warehouses than they have at Christmastime.”

Tsiounis also notes that the restaurant industry—which has been greatly hobbled due to closures and social distancing requirements—is down 89% year-over-year in the US.

“You know things are down, but you don’t know how down they are—are they down 50%, 75%, 25%?” he says.

A Hard Dig

Octavio Marenzi, co-founder and CEO of research firm Opimas, says that geolocation data needs to be used as “a sanity check” on consumer and credit card data, as the two have a lot of noise, and only include about 1% to 2% of all the credit card transactions in the US.

In 2018, Marenzi wrote a report on the geolocation space. Back then, he estimated that by 2020, asset managers would spend about $250 million annually on mobile geolocation data. Marenzi today says that estimate was a little high.

“The thing with geolocation data is it does require a lot of calibration,” he says. “It seems straightforward to say, ‘I’d like to count how many people were in this location at this time, and I’d like to know how that’s changing.’ But when you start to do it, it becomes quite complicated because you’re collecting the data based on app usage.

“While there’s a lot of noise, credit card and payment data are easier-to-read signals in terms of spending,” he says.

Geolocation is one of those alternative datasets that has a fair amount of proponents and detractors. There are several hurdles to overcome before this data can be remotely useful. It’s important to keep in mind that some phone operating systems and apps give location updates far more frequently (or infrequently, depending on how you look at it) than other operating systems and apps.

Additionally, cellphone tower triangulation is spotty, which affects GPS data. While GPS is the most accurate geolocation tracker, locating a device within three meters (almost 10 feet) can be impacted by terrain, buildings, and lack of service, according to Barr. And for satellite data, he says “the rule of thumb is that you use four out of six satellites to form a statistically reliable data panel.”

As noted before, no one dataset is likely to deliver alpha; there’s a blending with other datasets that needs to occur.

“Hedge funds, and especially quantitative hedge funds, are focused on point-in-time, time-series analysis. The first thing that you’ll see when you look at this data is that there are big gaps,” Barr says. “So if a hedge fund is using the raw data from the source—from the telcos or satellite imagery companies—the first thing they have to do is an enormous amount of data manipulation and cleansing in order to make sense of what the data is trying to say.”

Lastly, there’s a communication barrier that exists: oftentimes, hedge funds are not clear in what exactly they’re looking for. So you end up with a sea of data points, but no clear strategic direction.

“It’s challenging,” he says. “You have to know what you’re doing.”