My stumble on COVID-19
A couple weeks ago, I started investigating the response, here and in the stock market, to COVID-19. I found that LessWrong’s conversation took off about a week after the stock market started to crash. Given what we knew prior about COVID-19 prior to Feb. 20th, when the market first started to decline, I felt that the stock market’s reaction was delayed. And of course, there’d been plenty of criticism of the response of experts and governments. But I was playing catch-up. I certainly was not screaming about COVID-19 until well after that time.
Today, I found the most detailed timeline I’ve seen of confirmed cases around the world. It goes day by day and country by country, from Jan. 13th to the end of March.
That timeline shows that Feb. 21st was the first date when at least 3 countries besides China had 10+ new confirmed cases in a single day (Japan, South Korea, Italy, and Iran).
That changes my interpretation of the stock market crash dramatically. Investors weren’t failing to synthesize the early information or waiting for someone to yell “fire!” They were waiting to see confirmed international community spread, rather than just a few cases popping up here and there. Once they saw that early evidence, the sell-off began, and it continued in tandem, day by day, with the evidence of community spread in new countries and the exponential growth of COVID-19 cases in countries where it was already established.
Scott Aaronson speculates that he might have been able to tell what COVID-19 would turn into on Feb. 4th. I believed and agreed with him, hypothesizing that it was a lack of intellectual guideposts for synthesizing early information on a novel disease that made it hard to assess and act appropriately. To correct that lack, I created a model for what to lookfor and how to interpret information to judge the potential spread of a disease—an alarm bell for the next pandemic.
Now that I’ve found day-by-day, country-by-country confirmed case data, I need to change a few of my points of view.
First, is lack of confirmed international community spread generally a good reason to not panic, even if in this case things turned out badly? COVID-19 had its first confirmed case in the USA on Jan. 21st, but never had more than 1-2 new cases on any given day until Feb. 29th—with several days up to 12 days in between new cases during that time. That’s over a month of sporadic, occasional new case reports.
The coincidence of faster spread of COVID-19 internationally and the beginnings of the stock market crash suggest that hard evidence of community spread of a deadly disease is the world’s evidential benchmark for an economic downturn.
And that seems perfectly reasonable to me.
To argue otherwise is to say that it should be obvious COVID-19 was highly contagious based on some other form of evidence. That’s not impossible.
But in my case, I didn’t know until weeks after I started researching that the world’s largest encyclopedia had day-by-day, country-by-country confirmed case data.
So it’s really not a stretch for an amateur such as myself to have missed or misinterpreted some critically important source of data. I still think it’s possible for an avid, early COVID-19 amateur researcher to have felt reasonably confident in shorting the market prior to Feb. 20th. But they’d need to have searched long and hard for other forms of data to back up their argument. Not only evaluating COVID-19 on its own terms, but comparing it with other pandemics of the late 20th and early 21st century.
So I have to downgrade my confidence in the usefulness of the models I wrote for this. I need to call myself out on my willingness to assume with such confidence that the world had it wrong, rather than me. Maybe this experience has been necessary for my intellectual growth. Doing this research out of curiosity and being willing to accept my own conclusions, even when I ultimately felt I was mostly wrong, was perhaps the only way to get to this point of understanding, to get less wrong.
On a deeper level, I am learning and relearning a fundamental lesson lately. I don’t anticipate that I’ll have grasped it until it’s bashed me over the head at least a few more times. Here it is, in free verse:
Intellectual knowledge is a slippery thing.
I can’t readily grasp it.
I don’t practice engaging with it every waking moment, the way I do with my five senses.
Human intellect is 2% of the evolutionary age of the nervous system.
This kind of intellectual work is extremely difficult; only a few people care about it enough to make the attempt, so I’m getting far less feedback than I would for other activities.
I am no savant.
Others have been doing this full-time for decades longer than I’ve spent as a part-time amateur.
A major league baseball player who hits a home run has far more in common with an MLB player who hits a foul ball than a child who hits a home run.
Reliable knowledge is a hard-won and precious thing.
As a fellow amateur, I greatly appreciate your recent posts trying to understand the events of the last few months, as well as your willingness to publicly and humbly acknowledge your ongoing uncertainty and to update your beliefs based on new evidence. I feel that you may be being too hard on yourself here though—some comments below.
As you point out, the onset of international community spread does seem to be the tipping point for the stock market crash. But the possibility that it might spread beyond China/East Asia must have been apparent throughout Jan and early Feb. Why was this not reflected in market volatility? VIX also stayed very flat until Feb. 20, the same time the markets started to decline. Wei Dai pointed out that this was a contributing factor to the gains on his options trade. I could see stocks themselves remaining high until there was evidence of international community spread, but VIX/option prices at the time suggest markets failed to price in any substantial chance of this happening until it had actually happened.
I think there was a lot of evidence COVID-19 was highly contagious based on the reports from China. What was unclear at the time was 1. if China could control the initial outbreak successfully, and 2. if a few cases did spread, could other governments identify and quarantine incoming cases to prevent escalation to full community spread. Markets may have greatly underestimated the chances of these two scenarios playing out the way they did.
Totally agree. Predicting is hard, and predictions often look foolish with the benefit of hindsight. In this case, even understanding the past with the benefit of hindsight is challenging! But still worth doing, to try to understand the world better (or at least less wrongly).
True—but the only way for a child to reach the majors is through continued practice. Keep swinging!
Thanks for the encouragement! Here’s my take, best enjoyed with salt :)
It’s possible markets underestimated the chances of COVID-19 becoming the crisis it is today. But it’s also possible that they correctly estimated the chances at every step, and this time we got unlucky.
Looking at VIX back only 3 months takes you to mid-January—right at the start of the outbreak. Look back 6 months, and you’ll see that there is a marked increase in VIX starting around Jan. 23rd. Looking at the day-by-day cases, it’s right around Jan. 23-25th that you can see a market increase in the number of countries confirming new COVID-19 cases on a daily basis.
So my interpretation is that this jump in VIX represents the market’s probabilistic forecast that COVID-19 would turn into a major economic downturn. Investors were saying “Yikes, COVID-19 looks scary. But we’ve been through pandemics before and gotten them under control. Hopefully the same will happen here, but there’s a chance we won’t.”
It was around Feb. 21st that the crucial evidence of confirmed international community spread came in, leading to the VIXplosion after that date.
I still think the model I crafted gives some good guideposts for what to look for in terms of transmissibility: spreading via respiratory droplets, a peak of contagiousness coinciding with onset of symptoms, a fast increase in cases at the point of origin, and no widely available point-of-care rapid test. Together, these indicate a virus that spreads efficiently and is hard to screen for. If it’s deadly, too, then watch out!
But I don’t know how to evaluate whether or not the amount of S&P 500 30-day option volatility (which I understand is what VIX measures) during Jan. 23-Feb. 20 was too much or too little given the information we had. It seems clear there was some volatility—the markets didn’t ignore COVID-19 during that period.
Before, I believed my model can help users be smarter than the experts. Now, I believe it can help them approach being as smart as the experts. The key difference is that the former would help you beat the market. The latter only keeps you from losing your shirt!
Would also like to add that there was a significant knowledge gap in those early days in February, with data from China on both the scale of the outbreak, and circumstances, particularly difficult to interpret and extrapolate. Knowing what we know now, those early signs would have been enough, but at the time, it certainly seemed possible that it would be contained within China (very limited spread overseas that could be easily contained), or at least the mortality rate would be lower.
This is a better analysis of the stock market’s reaction than most others that I’ve seen.
The stock market was irrational in subtle ways, but didn’t make blatant mistakes like failing to understand exponential growth.
I expect the biggest problem was recency bias / availability bias that caused markets to overweight the chance that this virus would be contained in the way that happened with Ebola, SARS, and a few other scares that were the most familiar analogies in most investors memories. I think this was one of the larger mistakes that I made.
It’s also hard to evaluate the economic consequences of a given number of infections and deaths. The pandemics of 1918, 1957, and 1968 all seemed to suggest less economic damage than is being forecast for this pandemic. Some of that is simply due to us being better able to afford to shut down businesses than was possible for prior generations—restaurants and airline travel were a much smaller fraction of the economy in 1968 than today.
I think there’s a ton of complexity in “the stock market” that is pretty orthogonal to prediction of growth and impact of the virus. It’s very believable that this ends up BENEFITING the weighted majority of major indices—big companies with the political and financial connections to be in the index are exactly the ones that use their bailout money to buy the broken smaller companies on the cheap.
I don’t know enough to predict the changes, but it’s not unbelievable that we have large price and monetary inflation, even while the government lies about it with skewed measurements, so the stock market keeps up, and anyone who got out and put their hope in cash-equivalent or government-inflation-indexed instruments loses out.