In early February, forecasting tournaments, financial markets, journalists, random public health people, and amateurs all gave a low likelihood to a serious SARS-CoV-2 pandemic (<5%), and in hindsight, this shows it was legitimately hard to predict.
But we should not give up on better prediction, because our performance here was just surprisingly and unnecessarily bad.
These groups could have and should have given a much higher probability to what happened happening. Even, or especially, given the little we knew at the time.
My initial off the cuff probabilities gave a ~20% chance of it being wiped out like SARS1. We don't really understand how that could happen, but we also don't fully understand how SARS1 was eliminated, so uncertainty means there has to be a decent shot that the same will happen again.
This leaves an 80% probability that it will spread widely, which means there's a decent chance it will go on to cause millions or tens of millions of deaths.
What magic ability did I use to gain this insight which others missed? I formed a super simple inside view of what was going on, and bothered to use it.
That is: The virus moves from one person to the next via social contact. Lots of people already seem to have it. Given that, and high rates of travel into and out of Hubei in January, could we block off all of the current chains of transmission before it gets somewhere without the capacity to notice it or stop the spread? Probably not.
Here are all the routes via which we knew SARS-CoV-2 could become a dangerous global pandemic:
We've underestimated how many people already have it, and the true number is much more than my best guess at the time of 100,000. This would make it far harder to stop every chain of transmission.
We've gotten the start date wrong and it has been circulating since October or November. In that case it has likely made it to other countries some time ago, and begun spreading under the radar.
We've gotten unlucky and someone leaving Hubei has carried it anywhere else in the world, and passed it to a lot of people, perhaps via a super-spreader event.
It has already reached a few people in India, Ethiopia, Nigeria, or any other similar country. They won't have the infrastructure to notice it early enough, or wipe it out when they do notice it. And it will be very hard to cut off those countries from everywhere else, so sooner or later it spreads globally.
It's able to infect people with very low levels of exposure, or is partially aerosolised. In that case cutting off every transmission chain will be super hard.
Other developed countries won't do much to stop it from spreading after it arrives in their countries.
It spreads before people are symptomatic, so you can't practically isolate them before they pass it on.
Someone carries it out of China to a neighbouring and less organised country (e.g. Myanmar), despite the border closures.
China will not follow through on its lockdown efforts in most of the country, and will accept it becoming endemic.
China has been greatly misleading us about the origin of the disease, or its scope within China, or what they know about how far it has spread.
That list shows that there are lots of plausible and independent ways that a harmful global pandemic could happen, and the prediction come true.
(There are also a similar set of ways it could not spread widely, or not kill many people if it does, but I'll omit them for brevity.)
In a situation of such uncertainty, with so many ways the disease could escape control, it makes little sense to assign a <5% chance to it not causing a serious global pandemic.
As it turns out I was wrong about China not being able to control it internally. But it also turned out, i) SARS2 was pretty good at asymptomatic spread, ii) it was already in many countries by that point, and iii) most developed countries mounted no useful response early on. Any of these alone would likely have been enough for a serious global pandemic to result.
The mechanisms operating here were not so mysterious we need to rely primarily on a general 'outside view' to see the future. But if we did, a single success with SARS1 should not have been that reassuring, relative to our failure to control almost all (maybe any?) human-transmissible respiratory viruses — something so difficult we've almost never bothered to try.
My simple-as-hell inside view was consistent with opinion among actual experts in new respiratory pandemics. Neil Ferguson on 4th February (
(In future I will pay a bit more attention to people with relevant expertise rather than random officials in the general area like someone from the CDC, or a doctor who studied public health but not respiratory viruses specifically. And I'm downgrading my reliance on financial markets, which seem to pay little attention to real-world non-business issues their participants don't understand, like tail risks, including pandemics.)
Saying the risk of a major global pandemic was 5% or lower was a strange contrarian bet against i) the really obvious, and ii) expert opinion, more narrowly defined. And as far as I could see it was a contrarian bet that nobody at the time was trying to justify.
What was everyone thinking? I don't know. Some possibilities are i) pinning too strongly to the recent examples of H1N1 and SARS1 not being so bad, ii) just not paying attention, as there's lots of things to worry about in the world, iii) general skepticism that any one thing going on can be such a big deal, especially something so random and meaningless as a virus jumping from cats to people.
I'm sympathetic, especially to the fact that there's always so many damn things going on that nobody has the capacity to form a sensible view about all of them.
But I maintain that our forecasters did worse than they realistically could have, people should learn from this and improve their thinking, and we should aspire to give the world more sensible credences and forewarning next time.