Coronavirus & the Information Cycle in Markets
It's not easy training children to say "please" and "thank you" when appropriate (I must have reminded mine 10,000 times), and one unexpected benefit of the coronavirus pandemic is that my children now have a healthy hand-washing habit. Nothing focuses the mind like contagious disease.
Today's newsletter focuses on information contagion - how information infects markets, provokes denial and fear, and creates investable patterns during negative events like disease outbreaks. A PDF presentation with graphics and more information underlying the newsletter is available here.
For excellent alternative perspectives, see This New York times review projecting the disease burden of the pandemic. This James Mackintosh article in the Wall Street Journal discusses financial markets scenarios (e.g., L-U-or V-shaped bottom).
No one can deny the epidemic anymore - it's time to wash our hands and dig in.
UNDERREACTION (a.k.a. DENIAL)
I've given webinars and conference presentations on coronavirus over the past month (here's one), and I've been surprised by the wide range of attitudes towards the pandemic in different countries. As recently as last Friday I was on full flights in Scandinavia and occasionally shaking hands during meetings (and then nervously stuffing my hand in my pockets so I wouldn't touch my face).
However, at some banks in Stockholm meetings were canceled to prevent outsiders from bringing the virus into the firms. In other firms, in order to keep trading operations normal through the pandemic, employees were sorted into Team A and Team B alternating weekly home/work shifts. The general consensus in my meetings was that EU and US governments weren't taking the epidemic seriously enough. They were in DENIAL. As they say...
“Denial ain’t just a river in Egypt.”
~ Provenance unclear, but formalized by self-help guru Stuart Smalley (played by Al Franken), 1991.
Denial is how humans deal with surprising (usually unpleasant) new information. The academic term for denial in financial markets is underreaction. Financial markets initially underreact to both good and bad factual news. Note that I said "factual" news. Factual news is typically numerical and may be boring, routine, or repetitive.
In order to study this effect, we performed the study below. Using the past one-month of media sentiment, we see that stocks with the most positive media sentiment significantly outperform the most negative over the following 90+ days. The below study included 5,000 U.S. companies averaged over 2006-2018.
In this case, a computer can track (and predict) underreaction using our Refinitiv MarketPsych indices. (Graphic produced by our amazing quantiative researcher Anthony Luciani). This week we launched a new predictive stock ranking system with StarMine based on this effect, please reach out to us or StarMine for details on that product.
The initial bad news of the epidemic was not easily processed or understood, and it seems to have prompted a general sense of denial (at least in market prices in the US and Europe).
INFECTIOUS CURIOSITY
In countries with a recent history of impactful epidemics - like China/HK/Taiwan/Singapore with SARS - and after initial underreaction in China, COVD-19 was taken very seriously. In countries with little recent experience of a pandemic, and with significant uncertainty around the healthcare data coming out of China, the risks were generally ignored. This is denial.
Denial is very common when 1) times have been good and 2) available negative evidence is vague. To make up for the uncertainty of the data we rationalize to fill in the gaps, and we do it in a way that favors the status quo (called motivated reasoning): "it's like the flu, no big deal" or "the economic impact of panic wil be worse than a few nursing home residents dying, so we shouldn't do anything different."
Yet when others do the math, and in this New York Times review, the reality and projections (the potential facts) could be grim if serious preventive action is not taken. From a US healthcare perspective, if there is no full-scale effort to stop the spread:
1) COVID-19 is more contagious than the flu, and 34,000,000 Americans had the flu this year,
2) COVIDs mortality rate may be 1%, with 10% of sufferers seriously ill and needing hospitalization.
3) Unchecked and untreated 340,000 Americans could die from it this year, with up to 3,400,000 needing hospitalization.
4) There are 100,000 ICU beds in the US and until Friday, limited tests.
You get the idea - a surge in infections leads to major social and economic disruption.
Yet many investors seemed to know this was coming, and before the problems were evident in market prices, the discussions of them emerged in news and social media.
RISK-OFF SIGNALS WITH SENTIMENT
As we monitor media sentiment about all major assets, we see that media sentiment was declining well before prices began their recent tumble. This presentation shows slides of declines in media sentiment around the SP500, Crude Oil and Apple before the major selloffs. The case of Apple is below, from Sep 1 2018 through Mar 12, 2020.
The yellow bars are the Apple share price plotted alongside two moving averages of media sentiment (30 and 90 days). Green shading appears between the averages when the 30-day average is above the 90-day, and red shading appears when below (when recent sentiment is falling). Note that downturns in sentiment may happen before prices fall as in Oct 2018 and in Jan 2020. In Jan 2020, news and social media was very negative on Apple's outlook before the company itself warned of losses.
Sentiment is the media's perception and it forms as people discuss and process the meaning of recent events. With some events - like disease outbreaks - the event itself is a negative, and we can act immediately if we know the event is bad for the market.
RISK-OFF EVENT: HUMAN INFECTIOUS DISEASE OUTBREAK
When our quantiative researcher Tiago Teodoro looked at country stock markets and airlines in 2020, he found that the first spike in "Human Infectious Disease" (the sum of references to all diseases that affect humans) for a given country precedes multi-week drops in airline stocks of that country. We have many examples in this presentation including the chart below.
In the graphic below the light blue line represents the value of Japan Airlines stock through Mar 10, 2020. Yellow bars along the midline are the frequency of references to Human Infectious Disease in the media about Japan. The lower subplots depict Japanese Fear and Sentiment moving averages for the entire country.
Note that as of Monday March 9th, Fear was still rising and sentiment falling in Japan.
OVERREACTION
Once media fear catches up to (and surpasses) the negativity of the facts, we are in the realm of emotional overreaction (panic). That's when markets are panicky and volatile (e.g., Thursday's largest point drop since 1987).
In periods of high emotionality (where emotions dominate over facts) prices often move sharply in one direction and then reverse. Such V-shaped (and inverted V) patterns are called overreaction.
When we ran our social media-based quant fund in 2008-2010, we found that a week of high fear coupled with a week of declining share prices often preceded a 4-5 day bounce. That is, it's good to buy on fear, but only for a week. After a week-long bounce investors again re-evaluate the facts. If bad factual news is growing (e.g., travel & conference cancellations, death or infection of more prominent people), prices head lower again.
What ultimately triggers a sustained bounce is a conviction that the problem is understood and will be addressed. With weak leadership, as was feared after Trump's speech on Wednesday night this week, investors catastrophized about a tsunami of victims flooding hospitals and severe economic contraction. With leadership, preparation, and planning, the risk is mitigated and future projections look manageable (as investors took away from Trump's State of Emergency declaration on Friday).
As Dr. Anthony Fauci noted in a U.S. Congressional hearing: “All models are as good as the assumptions that you put into the model....What will determine the ultimate number ... will be how you respond to it with containment and mitigation.” If containment and mitigation is pursued in earnest, the economic impact may be lessened and the economy can bounce back.
THE BOUNCE
In Tiago's research on airlines, when both fear declines and sentiment starts to improve, then the longer-term bottom is (often) in place. See the various airlines examples in the presentation. One seasoned trader I met (and shook hands with) in Oslo last week said he finds the bottom when the news continues to get worse, but prices have stopped falling (an inverse "news failure").
Bottoms seem to occur when sentiment is still very low, but prices stabilize for more than a week. We don't have a model to identify this relationship yet, but we're working on it.
One of our clients tells me that the "uncertainty collapse" when a solution to the crisis is at hand is itself the marker of a bottom. Will the concerted action by governments this week be the trigger for that uncertainty collapse? Short-term relief certainly, but it depends on whether the perception that the virus is out of control resumes.
One academic paper on my desk is titled "Relief from incidental fear evokes exuberant risk taking." That title explains why so many researchers are trying to time the market bottom - it can be very lucrative if timed before investors exuberantly buy back into stocks. When they re-enter stocks en masse, the bounce will be large and swift.
In this presentation (and last month's newsletter) we describe examples of bottom-fishing in past epidemics. Both China and Hong Kong appear to have bottomed based on Fear and Sentiment trends, as you can see both Fear falling and Sentiment rising around China (below graphic). But those signals were early. And media about China is shaped by the government's messaging. In the chart below, notice how references to viral pneumonia (human infectious disease) in the Chinese media spike in late December, fall off in the second week of January, and then surge upwards again around Jan 17th - possibly reflecting government censoring of viral pneumonia reports.
In the graphic below, the light blue line represents the value of China Southern Airlines stock through Mar 4, 2020. Yellow bars along the bottom are the frequency of references to human infectious diseases in the media about China. The lower subplots depict Chinese Fear and Sentiment moving averages.
China's massive liquidity injections into the stock market held prices up artificially, so those market prices aren't naturally evolving. Yet in China people are returning to work, and 80% of Apple stores have re-opened, and it appears the ground-level reality is stabilized with a manageable rate of disease spread. This is likely the best that can be hoped for until a vaccine arrives.
WHERE ARE WE NOW?
The virus is more contained in some countries (China, HK, Singapore, Japan) than others (the EU and the US). For bottoms to become evident, we need to see stronger government responses in infection control to stop the spread in the EU and the US. "When the US coughs, the world catches cold" is true of both China and the US now. Trump's emergency declaration on Friday may have been the action the market was waiting to see. Of course, this is all speculative - markets (like diseases) are not simple. Please see Disclaimers below.
HOUSEKEEPING AND CLOSING
I'm sometimes asked if I personally "eat my own dog food" - that is, do I trade on these signals. Frankly, I find it distracting, and I try not to time the markets unless I have a powerful hunch (maybe once per two years). I thought this epidemic was serious, and I eliminated half (but not all) of my equity exposure in mid-February. So yes, in this case I did take action. But usually I don't.
There are values emerging in many markets, and if you can buy a company with more cash than its market value, that's usually a good margin of safety and a promising long term investment (and a few such opportunities emerged on Thursday).
There are also new risks in the functioning of markets as some strategies implode or are unwound, and it will be interesting to learn who was "naked when the tide went out."
Promises to be an interesting Spring. A big bounce may be coming next week, but could easily unravel again after a week. Ultimately we see some real bottoming coming by April (if not sooner) - but this is fast-moving, and the data isn't giving us a bottom signal just yet. Please get some sleep, stay safe, and stay healthy.
Happy Investing! (And wash those hands!)
Richard and the MarketPsych Data Team