Peter Hafez, Chief Data Scientist, RavenPack
| March 17, 2020
What light, if any, can news analytics shed on the coronavirus epidemic? In this post, Peter Hafez, Chief Data Scientist at RavenPack, relates his findings, including how, in many cases, a spike in news volume leads official reports of confirmed cases, and how virus-news as a percentage of total news can provide an indicator for market returns in sensitive sectors.
The coronavirus strain responsible for the COVID-19 pandemic has the world in a panic. The CDC has said that
the virus is unlikely to be containable
and epidemiologists estimate that it could eventually
infect 40-70% of the global population
. Investors, policy makers and the broader public are understandably trying to decipher what the final ramifications will be.
High levels of uncertainty and lack of detailed scientific data allows sentiment to drive market moves. The below chart shows that the recent spike in “Recession" chatter is fueled by the coronavirus: more than 80% of all news talking about a recession makes mention of the coronavirus, showing a clear linkage between the two topics and what is currently driving market fears.
It’s important to distinguish between the medical, psychological and economic implications of this outbreak. Investors are struggling to parse the rapidly incoming data, but we do know that the virus is highly infectious, with an
R0 estimate of 2.2
. Mortality rates vary widely depending on the source, but it’s looking to be higher on average than seasonal influenza, and higher in older cohorts and those with respiratory comorbidities. The
WHO estimate of the 3.4% mortality rate
likely demonstrates a large severity bias, and as broader, more rigorous diagnostic testing strategies get under way, we will get a better sense for the denominator and the true intrinsic lethality of the disease. The actual number of cases is likely much higher, as 80% of those infected suffer
minor cold-like symptoms
. This means that mortality would come in much lower, but would also result in the wider spread of the disease with people walking around with mild symptoms.
The below map shows the progression of confirmed COVID-19 cases, overlaid on top of relative news volume related to pandemics and epidemics by country (darker green = higher news volume, darker red circles = higher mortality rate). We can see that in many cases, a news volume spike leads the official reports of confirmed cases.
Map of epidemic/pandemic news volume and COVID-19 cases by country from Jan 2nd - Mar 12th 2020. Sources: RavenPack, March 2020; Johns Hopkins COVID-19 case data
The effects on human psychology are obviously quite strong, and it is unclear at this point whether the extreme responses that we’re seeing are warranted. Over the weekend, many countries have implemented drastic measures in an effort to contain the spread, such as closing down borders, banning large gatherings, instituting travel bans and closing down public places, as well as declaring nationwide states of emergency. What we know is that these types of high-anxiety events can result in outsized reactions from policy makers who come under high amounts of pressure to take some action, and that gradually taper off as the population acclimates. The extreme measures taken by China seem to have some containment effect, and as time passes we see a dropoff in the overall news volume related to pandemics and epidemics in the region, while the EU and the U.S lagged behind.
The concerns surrounding the COVID-19 pandemic trumps those of SARS, H1N1 and MERS. Advances in information technology allow for much better information propagation and enable the impressive rate at which scientists are able to collaborate on sequencing the virus and work on the vaccine. But this also feeds the spread of fear and misinformation across the abundance of online sources that didn’t exist during prior epidemics.
While the containment measures are unlikely to prevent the virus from spreading to 40-70% of the global population, they can help with flattening out the epidemic curve and relieving the pressure on the health care systems by
decreasing the intensity of the spread
. The above world map shows increased mortality rates in places where access to higher quality healthcare is limited and in locations with a higher concentration of an ageing population. While the 2004 SARS outbreak had a mortality rate of close to 10%, it never became a pandemic as this very fact contributed to it burning out relatively quickly. The 2009 H1N1 outbreak caused a broad panic, but fizzled quickly. COVID-19 seems to have a slightly higher fatality rate, but it remains to be seen whether it too fizzles out. The nightmare scenarios summon up the Spanish Flu of 1918, which tapered out on two occasions before it wreaked havoc on the global population, infecting 500+ million and killing tens of millions.
If we look at a normalized level of pandemic-related news volume over time, we can draw a better comparison between these large recent outbreaks.
With the increasing introduction of cordon sanitaire, business closures and social distancing across regions, the short-term impact on the economy will be severe, but the major question is what the longer-term implications will be and whether we will see a growth catch-up over the following quarters. Market volatility is a response to the uncertainty around the latter question and whether the true scenario is being discounted properly. A major shock to the credit system can cause serious adverse long-term ripple effects that destabilize the global economy. Policy responses and central bank liquidity injections can help, but might not have the needed size to sufficiently mitigate the shocks emanating from a worst-case scenario. Depending on how the situation evolves, it is helpful to know which companies are often co-mentioned with the disease by sector in order to identify where stresses might materialize. The below plot shows the top sectors with the highest number of co-mentions with previous epidemics.
An obvious casualty of the travel bans, which are taking a heavy toll on the travel industry, are airline stocks, having been hit with losses of over 40%. The following plot shows the relationship between the American and European airline companies’ exposure to China and the respective cumulative returns over the first week of the coronavirus outbreak. China-related company news as a percentage of total news is used as a proxy to infer the airlines’ exposure to the Chinese economy.
The relationship clearly shows that companies with greater exposure to China-related news initially got hit much harder than those with lower exposure. Airlines with lower exposure only started getting hit a month later when the outbreak began spreading globally.
On a more positive note, with the development of the COVID-19 vaccine on fast track, we found that the biotech companies with a higher co-mention volume with the virus were also the top performers in their sector. The following chart shows the relationship of the co-mention volume with cumulative excess returns over the entire period since January 14th.
We will keep monitoring the fluid COVID-19 situation very closely and provide regular updates on new approaches and ideas around tackling the incoming information flow. As we do this, it is important to consider the context of this rapid escalation in global developments. What’s clear is how poorly equipped the world is to deal with a virus that likely kills less than 1% of those afflicted. In contrast, certain types of avian flu, such as the highly lethal H5N1 strain, are estimated to have
fatality rates upwards of 50%
. Luckily, human-to-human transmissions are rare, but a mutation that would enable for a more efficient transmission mechanism would obviously be quite disastrous. The world can consider this pandemic a test run for a more serious challenge down the road.
Coronavius Data-Driven Insights Hub
where we are pulling together all our analytics, information and insights about the novel coronavirus in one place.
Please use your business email. If you don't have one, please email us at email@example.com.
We will process your personal data with the purpose of managing your personal account on
RavenPack and offering our services. You can exercise your rights of access, rectification,
erasure, restriction of processing, data portability and objection by emailing us at firstname.lastname@example.org. For more information, you can
Your request has been recorded and a team member will be in touch soon.
We consider incorporating sentiment signals from news, earnings call transcripts, and insider transactions to
boost the risk-adjusted returns, and revive factor performance.
We find stronger, more predictable market reactions when the words of company executives agree with their actions.
We have gathered 12 insights from 2021 research that can be leveraged in 2022.