1. Tyler Cowen on the Grand Princess data (not quite the complete sample you thought it was). UPDATE from a commenter: Diamond Princess was the complete sample. Grand Princess is the cruise in SF Bay.
Most of my rationalist friends self-isolated really early, before it was socially acceptable to do so
I cannot resist joking that nerds on the spectrum look for any excuse to self-isolate. But seriously, the people who are complaining about “panic” and “you’re trashing the economy for nothing” and “the flu is worse” strike me as less facile with math and data than those of us on the other side. Almost three weeks ago, Tyler Cowen talked about the debate between “growthers” and “base-raters.”
There are still base-raters out there. In a podcast interview with Nick Gillespie, Richard Epstein insists that outbreaks look geometric when they start, but then people adjust and they slow down. But I cannot think of a country where a slowdown occurred without strong government action. We’ve had effective government action in a few countries, and not-yet-effective government action in many others.
3. John Cochrane laments that economists are using old playbooks to deal with the current situation. I would say that there is a divide between the “stimulus” approaches and the “forbearance” approaches. He points out that the forbearance approach has not been carefully worked out. But that is not a good reason to prefer the stimulus approach. Doing the wrong thing because you know how to do it strikes me as the policy equivalent of looking for for the lost keys under the lamppost because that is where the light is.
Richard Epstein: ‘More Probable Than Not…Total Number of Deaths at Under 50,000’ on The Reason Interview With Nick Gillespie.
I listened to the Reason podcast with Richard Epstein yesterday. Epstein understand the problem with exponential models but he, in my opinion, does not have a good understanding of when exponential models are useful nor the output of a better model given the current scenario.
The exponential model is useful to estimate the early high growth phase of the real-world data which will end up looking like a Logistic Curve (with a kapacity K). It is useful because it is a single equation with one input variable (in the simplest case).
A good model will never be represented by a single equation; it is an algorithm not a calculus. The proper algorithm is incremental, performing a calculation one node at a time. It will draw from Graph Theory (like the Facebook Social Model), Timeseries data, Geospatial data, state machines, and plain Jane Relational Data (linked tables). The Logistic Curve (and its Gaussian derivative) are useful to judge the output of the algorithm rather than being the basis of the model itself. The Logistic/Gaussian curves are useful for eyeball heuristics of where we are in the growth but when there is no predictive power early in the curve, it always looks the same until growth slows. The key to this kind of model is that the calculation at each node/timestamp is based on a human interaction model.
When I criticize the exponential model I am not being optimistic about the outcomes, I am being pessimistic about our collective ability to focus our attention where it is most needed.
Over the last five or so days, the problem with exponential thinking is that it causes panic whenever we find a confirmed case with undetermined origin. The exponential modellers chime in and claim that this case represents hundreds or thousands of undetected cases. When carefully documented, these undetermined cases stop rather than explode. We have to trust the math and watch carefully for Node centric community spread every 5 days but we shouldn’t assume we have unconstrained community spread until this growth is sustained for 10 or more days. Either way, the Suppression strategy continues unchanged until Health System capacity is exceeded.
I again basically agree with everything you’re saying here, and particularly that the countries that have fared best have taken extreme actions. And I suspect they would see outbreaks accelerate again if they tried to resume normal activity.
Of all the analytical blogs I follow, I only see one that I respect still holding on to a view that this will not be a big deal. And even his forecast for total cases jumped 6X in one week after incorporating new data this week.
Diamond Princess was the complete sample. Grand Princess is the cruise in SF Bay.
And the Diamond Princess data is still unfolding, which is why I’ve mentioned it in the past without giving links. There were 250 Canadians originally symptom free passengers from the Diamond Princess quarantined at Canadian Forces Base (CFB) Trenton. There are new cases from this cohort still popping up in the 14-day isolation and these are included in the Ontario numbers (same is true for the 219 Wuhan evacuees that completed their quarantine at CFB Trenton around Feb 25). Some of these individuals only started exhibiting symptoms while in the air on their way to Canada. If I remember correctly, 14 of the 250 Diamond Princess Canadians subsequently tested positive sometime over their 14 day quarantine.
…the people who are complaining about “panic” and “you’re trashing the economy for nothing” and “the flu is worse” strike me as less facile with math and data than those of us on the other side.
Razib: “To be frank, most of the skeptics of the impact of coronavirus are not very smart.” https://www.gnxp.com/WordPress/2020/03/18/covid-19-and-the-sars-cov-2-pandemic-how-bad-will-it-get/?utm_source=dlvr.it&utm_medium=twitter&utm_campaign=covid-19-and-the-sars-cov-2-pandemic-how-bad-will-it-get
Ontario’s Chief Medical Officer of Health, Dr. David Williams talks for ~8min about exponential models at 18m40s of today’s Update on COVID-19 Response
I’m not optimistic about the virus. I’m pessimistic about our ability to quarantine well enough, long enough to make a difference. Food and medicine don’t come from nowhere; people have to go to work to make them. And unless a quarantine successfully isolates every single infected individual until the infection passes (very unlikely) and successfully prevents every single contagious individual from entering (also hugely unlikely), the pandemic will just restart when we can’t sustain the quarantine in a few months. Then everybody will get sick and we’ll be out of stored food and medicine.
Jay, I think you should be more optimistic. The Self-Monitoring and Self-Isolation policies really do work; it all comes down to cough droplets. It looks wildly uncertain now only because people were initially unfocused when it was an abstract foreign problem. The U.S. didn’t have the institutional memory and now it’s learning fast.
I’m not saying those policies don’t work. I’m saying they can’t be sustained indefinitely, and the problem will still be there when we can no longer sustain them.
I hear ya and I understood your points but I think the policies work faster than you assume once people are roughly pointing in the same direction.
The Epidemic YoYo is a concern but I’m sure the solution will be more obvious before that bridge has to be crossed.
You might be right. We can buy some time at the cost of stored resources; we need both and I don’t know how much time we can buy or when the trade turns net negative. South Korea was doing well with its test-and-quarantine system for a while, but it looks like the virus is getting away from them. Hard to say what’s best.
That is a great article, Jay, thanks. Everyone focuses on the widespread testing but they don’t mention how central Contact Tracing was to South Korea’s “success”. They also were in constant contact with those that tested positive. The real trick to this is getting people to change their ingrained behaviors.
The problem with these articles is that they get caught up in the flashy narrative and the easy, obvious, and mundane seem to escape us. Right now the U.S. and Canada are dealing with a Travel centric infection yet no one is jumping up and down and asking the obvious question of why aren’t we using flight and border data as the starting point of the giant call center effort that is going to be required. No one is asking how many Americans own a thermometer to track fevers.
Highly recommend the Joe Rogan show with Michael Osterholm. M.D., Ph.D., an internationally recognized expert in infectious disease epidemiology. He is Regents Professor, McKnight Presidential Endowed Chair in Public Health, the director of the Center for Infectious Disease Research and Policy (CIDRAP) on March 10th.
Harsh truths but from someone who knows and doesn’t favor lying to the public.