Big data is already on the front lines of the fight against coronavirus. China’s authorities are using advanced technologies like thermal imaging and facial recognition to track infections and enforce quarantines. Now investors are putting similar technologies to work to analyze the financial impact of the epidemic.

Coronavirus fears caused markets to lurch more violently than at any time since the Lehman meltdown. In response, some firms are digging deep into data, unearthing non-traditional data sources—from air quality indexes to traffic jam counts—in attempts to forecast possible losses. Or even turn a profit.

New data can shed light on how fast countries are getting back to work as the epidemic recedes and can identify which companies are hurting—or benefitting—from the fallout. Much of the effort is focused on China, but the principles can be applied elsewhere as the virus moves on.

So-called alternative data is timelier and can be more reliable than official statistics. It can fill the void created by the lack of comparable events. Alt data will allow investors to build an understanding of coronavirus’s local effects and extrapolate from there.

It can be a delicate subject to broach, however. One practitioner describes coronavirus as the “perfect application” for the new data science—but is at pains to emphasize the global pandemic was not welcome.

And gears are grinding to get alt data in motion across the industry. JP Morgan, Morgan Stanley and UBS are among the sell-side shops employing alt data to get a quicker and more accurate picture of the virus’s impact.

On the buy side, AllianceBernstein has set its technology to monitor what different companies have to say on the virus. “We realized there was going to be a peak in cases outside China and the economic impact was going to be a lot larger than people had thought—and we wanted to understand what European and US corporates were saying,” says Andrew Chin, AB’s chief risk officer.

Other asset managers also report working on alt data applications but chose not to be named for this article.

First test

The coronavirus represents the first big test for alt data’s ability to help investors during a global, macro market-shaping event. It has come into play during previous crisis scenarios, but on a smaller scale.

After the Mumbai terrorist attack in 2008, some analysts gathered information from rickshaw drivers about the number of foreigners they were carting around the city in order to get a handle on the attack’s economic impact.

Today, firms like UBS can collect ground-level data from thousands of sources to “swarm” an area of investor uncertainty, says Barry Hurewitz, head of UBS Evidence Lab, which produces research using new data sources.

Buy-side firms such as Schroders and Goldman Sachs Asset Management have established teams to gather and track new sources of information. Other banks have reinforced research functions with the addition of data scientists following the Evidence Lab model.

Much of the work happening now aims to track how fast China’s workers are returning to their jobs after enforced factory closures and an extended Lunar New Year holiday.

JP Morgan’s China equities team has incorporated air quality data into its research to assess how industrial production is rebooting after shutdowns.

Several firms, including JP Morgan, Morgan Stanley and UBS, are using traffic data from China’s Baidu internet service provider—akin to Google Maps—to determine the rate at which employees are returning to work.

The airline industry has never seen such a dramatic reduction in airline capacity from one event

Ronan Crosson, Eagle Alpha

Other data is helping gauge the impact of the epidemic at the company level and to identify the most exposed industries.

Airline capacity data—sourced from the airlines themselves—shows a far greater drop in travel than after the September 2001 terror attacks or Sars outbreak in 2003.

“The airline industry has never seen such a dramatic reduction in airline capacity from one event. As of mid-February, the industry had lost 10 million seats to, from and within China, which could result in a revenue shortfall of nearly $2 billion,” says Ronan Crosson, director of data strategy and analytics at Eagle Alpha, a broker of alt data.

Buy- and sell-side firms alike are using natural language processing (NLP) algorithms to trawl for insights into the epidemic’s impact on companies.

AllianceBernstein set its algos to search company releases for mentions of coronavirus in mid-February to spot clues on sales or earnings trends. “An algorithm can do this much faster than analysts can and a lot more comprehensively,” says Chin.

The new data can provide targeted insights. At investment manager Havelock, chief executive Matthew Beddall says the firm may use data on cruise ship moorings to gauge the industry’s performance.

Online video services are among the companies that appear to be benefitting from the crisis, with purchase receipt data showing a 30% growth in subscriptions in China in the past month, says Crosson.

Tracking mobile app downloads, UBS saw the use of online schooling services double in the month to February 9. It also used geolocation data to work out which fast food chain was most prevalent in Hubai province.

Tracking the movement of app users, an Eagle Alpha webinar reported a fall-off in trips to Nike stores in China—and to Macau casinos.

Tracking anxiety

Certain firms are trying with new data to map company supply chains, something  Goldman Sachs Asset Management has worked on.

Dun & Bradstreet, the business data company, is pitching to investors a database of billing relationships between entities that shows how businesses are linked to others in afflicted areas. Some 51,000 companies around the world have a direct relationship with a supplier in China’s Hubei province, the epidemic’s epicentre, the firm says.

One vendor, analyzing web data by NLP, identified 69 companies that had suspended operations as a result of the virus outbreak. 

New data can help also to gauge market anxiety, some think.

Quants at Wolfe Research have constructed an index of coronavirus sentiment using NLP to analyze news stories about the epidemic.

Wolfe looked at the sensitivity of different sectors in China to worsening Covid-19 sentiment and found energy, capital goods, transport, retailing and real estate to be the most exposed.

Conversely, real estate was one of the less vulnerable sectors in the US, due to the more defensive nature of the sector, which is dominated by real estate investment trusts, says head of quantitative research Yin Luo.

Using alternative data from China could also serve as guidance for what’s likely to happen in Europe and the US as the outbreak spreads.

Hurewitz describes this as building “real-time base rates”—benchmarks against which statistics from other regions or periods can be cross-checked. He compares the epidemic to other natural catastrophes that alternative data can render into data points as the events unfold.

In a tropical storm, Evidence Lab tracks rainfall by zip code, for example, building a live picture of the storm unfolding. It can then track its economic impact through data from transaction receipts or data on traffic levels.

“When there are no established patterns, when an event is rare, you need to understand what’s going on in real-time,” says Hurewitz. “Alt data is one of the few mechanisms with which you can start to establish reference points so as to triangulate the data and figure out what it means.”

And what the data disproves could be as valuable as its revelations.

Analysts armed with a cloud of facts are better able to challenge the hunches that dominate investor behavior in panicked markets, such as the idea that coronavirus spreads more slowly in hot climates—which, so far, is uncertain.

Many investors are using models based on the Sars epidemic to gauge the likely impact of coronavirus. But alt data says the epidemic will likely follow a different pattern, not least because travel times have halved across China since 2003, making for a more connected population.

Fast and fair

Alt data is seen as especially useful in determining activity in China where many investors mistrust official statistics—including reported numbers of coronavirus cases.

At Havelock, which is keeping an “informal” watch on air quality data as a measure of industrial activity in China, Beddall says alt data reported by US embassies and consulates is potentially more reliable than numbers from China officials, which could be biased.

Alt data is also fast data. Maurizio Luisi, a quant who built a nowcasting apparatus at Goldman Sachs, and now heads research at alt data-driven nowcaster SquareMacro, says China’s inflation indicators spiked by one standard deviation at the beginning of February. Such a big move pointed to a “major disruption”, Luisi says, well ahead of the global market sell-off.

Metrics of consumer and business sentiment, based on hundreds of surveys collected by SquareMacro’s partners in China, plunged an “astonishing” eight standard deviations last week, he adds.

In the seven years since its inception, Eagle Alpha has seen nothing comparable to recent events, says Crosson. “Every so often there’s a spike in interest in Tesla or Bitcoin or something like that. We’ve never seen such a sustained spike in interest and in a single topic.”

When there are no established patterns, when an event is rare, you need to understand what’s going on in real time

Barry Hurewitz, UBS Evidence Lab

Around 50 buy-siders joined the company’s webinar last week in which 11 vendors touted datasets showing the impact of the epidemic using data from satellite imagery to airline ticket sales or subscription rates for video-streaming services, such as iQiyi and Tencent.

Rado Lipus, founder at broker Neudata, says a trickle of coronavirus-related inquiries started in February and in the last week nine out of 10 inquiries have concerned the growing epidemic.

The Evidence Lab team at UBS has organized a “swat team” of senior managers to gather up existing data that might be relevant and commission new hunts for data it expects to be in demand. Coronavirus is now the number one inquiry among the lab’s data users.

Across the industry, rapid mobilization has tracked the speed of events. Coronavirus emerged in China’s Hubei province in December last year, infecting more than 96,000 people and killing over 3,300 at the time of writing.

The spread of the virus to other countries triggered an 11% fall in the MSCI World index in the week of February 24, only exceeded in the past 50 years by the aftermath of Lehman’s 2008 collapse and in 1987 by Black Monday.

Economists are sketching a hazy picture of the likely effects on the real economy. Consensus forecasts predict a sharp dip in China's GDP in the first quarter, but a recovery in Q2 and a modest 0.2% downward revision of forecasts for the year. Those projections, though, come laden with qualifiers.

China’s purchasing managers index sank to an all-time low in February, implying Q1 GDP could be as much as 5% lower than the last quarter of 2019, Wolfe Research analysts estimate.

Shifting perceptions

On the buy side, it’s easier, for now, to find those eyeing alt data with interest than to find firms using it to make trading calls.

This is partly because of the speed at which the coronavirus epidemic has unfolded. The chief quant at one leading asset manager describes the situation simply as “pure uncertainty”. Havelock’s Beddall calls it a “massive unknown”.

But investment managers agree, alt data has a growing role to play, particularly for firms investing on shorter horizons. “I can see how firms focused on next quarter’s earnings for companies can get some juice out of this,” Beddall says.

Chin says AB’s insights into managers’ sentiment on the virus mainly expose uncertainty at this stage but will become more meaningful in time. “Next month we just have to press a button and out comes what companies have been saying, how that’s different from January and February, and so on. That’s how we expect to use this.”

As for what the broader data says now, coal consumption by power stations in China has reached 62% of pre-Lunar New Year levels, compared with about 90% in other years. Passenger traffic and intra-city congestion suggests workers are trickling back to their posts. Congestion has edged up to 83% of the comparable level this time last year, Morgan Stanley reports.

The data on coronavirus sentiment is limited because it relates only to recent weeks, warns Luo, but it seems to be saying to go long China stocks, particularly beaten-down stocks in domestically focused companies.

The picture in Europe and the US is different, he says, as uncertainty rises and sentiment deteriorates. “There is probably more upside in China,” he suggests.

“China may yet emerge as a winner rather than a loser from this, compared with Europe, where arguably the risk of the epidemic spreading is bigger,” says Luo.

Since the end of the enforced holiday on February 10, although air pollution in China has increased, it remains at about 50% of the historical average.

Every cloud…