Friday, April 17, 2020

Debate Over Covid Fatality

So, first came this horrendous WSJ article by a Stanford couple claiming the true Coronavirus death rate may be much less than what people think. They suggest that perhaps the real rate may be as low as .01%, and thus, that maybe we should just let the disease run. After all, if just between 20,000 and 40,000 Americans die, what is the harm? 

Of course, now, with 37,000 Americans already dead from Covid-19, and the body counts still rising daily (some 2,500 just yesterday), this estimate now looks foolish. Not to say, "I told you so", but the WSJ looked problematic to me in real time, so I tweeted about it:

In retrospect, my response looks prescient, in part since China's own government revised up its estimates of death in Wuhan.

Debraj Ray, an economist who I respect a lot, and who is an editor of the American Economic Review, also posted some interesting thoughts about the Covid-19 death rate here, influenced by the Stanford couple's WSJ article. He argued we should rethink lockdowns given the uncertainty over the fatality rate. I posted a response on his blog, which I'll repost here:

Hello Debraj. Thought provoking post. I see some potential problems in your logic, however. Why are you ignoring the growth in deaths (or lack thereof) in New York before March 14th? If it's true that there were no deaths before then, then why wouldn't it follow that cases before then were scarce? Of course, likely there were few deaths before March 14th because there was little testing -- if you die of flu symptoms with no test for Covid available, it's likely you are not marked as dying of coronavirus. To the extent this is true, the growth rate of deaths is slower that what you have merely assumed. And, if it turns out that the rate of growth doubles every four days, then that would imply that either New York had it's first case before Wuhan did, or the death rate is even higher than 1%.

Lastly, Bhattarchaya's stuff has already been debunked. The death rate in Vo was 1%. The Chinese data should not be taken at face value. The NBA is not a random sample of the US. The Stanford couple mention .01% as a plausible death rate. But, some Italian provinces have death rates already as high as .065%, and that is assuming a 100% infection rate, which can't be right, as these provinces are still recording masses of new cases despite a draconian quarantine. In other Italian towns, already .1% of the population has died, and many people have also tested negative. 

After all this, I should mention that I'm actually in agreement that the death rate could well be much less than 1%. .4-.6% seems plausible to me, but spreads easier since there is no vaccine. On the other hand, .1% seems ruled out by what we know already. This could explain why Italian hospitals were overwhelmed so early, why they are not by the flu. 

Also, you write that people have freaked out too much. Have you paid any attention, at all, to what has happened in Italy? The freak out there has not prevented ICU units to become completely overwhelmed. They are essentially letting people over the age of 65 die for lack of ventilators. You really think this mass "freaking out" in Italy makes no sense?

Very Respectfully Submitted,Doug Campbell

Debraj then provided a thoughtful response.

Hi Doug, thanks. I wish we knew the truth behind fatality rates. You are right that it could be as high as 0.4%. My post is not written with a sense of dramatic certainty. That said (1) could you give me references to the Vo fatality rate? (2) I don't know what you mean by trusting the "Chinese data" --- the data has to do with airlifts out of Wuhan where everyone was tested, (3) the flu rate is for a particular distribution of ages (say, the US distribution), and we need to correct for that age distribution when studying regions with a high proportion of elderly, and (4) please read carefully before commenting on my interpretation. My post is perfectly consistent with OVERALL deaths being higher by a factor of 3 or more. I don't know what you want to include under the term "freaking out," but I am not including the terrible sense of sorrow that people feel for losing their loved ones. I simply refer to one's *own* fear of being fatally sick. Thanks.

He agrees that a death rate as high as .4% is plausible.  In response to his point:

    (1) I actually don't see the Vo fatility rate that I had seen earlier when I was just googling, although they had like 89 people test positive when they tested the first time, and they had a high-profile early death:

   (2) Bhattarcharya's first piece of evidence, using Chinese data, takes the fatality rate in Wuhan at face value, while making an adjustment to the infection rate. I was suspicious of the Chinese fatality rate. Lo and behold, under pressure, China has now revised up its fatality estimates for Wuhan by 50% -- and many people, me included, suspect it even could be higher than that. I won't try to claim I'm any sort of a genius here -- it was obvious the Chinese data were fake. Still is. It was troubling to see the Stanford duo take them at face value.

   (3) I agree we need to adjust for ages. Also, in Italy, Wuhan, and Iran, the hospitals were clearly overwhelmed. But, the death rate in Lombardy is now at .12%. Over two-thirds of the people there tested, tested negative, and they only test high-risk people. In addition, they have undoubtedly missed a lot of deaths, and people are still dying. You add all this to the info that their hospitals were overwhelmed by the end of February, and it's clear this is vastly more deadly than the flu. You had written: "That calls for a fatality rate around 0.1%". Even in New York City, the death rate is already around .1% of the population (that is, % of the population before the exodus), and over half of the people tested there have tested negative. And this came in the US, without any overwhelmed hospitals and likely a similar population structure, if healthier and wealthier than average. Even if we assume that 20% of the New York City population has already been infected, then we get a fatality rate of around .5%. If 25%, then .4%. Given the exodus of people out of the city, and 60% negative test rate of those with severe symptoms or direct exposure, it's probably tough to make an argument that more than 25% of NYC residents could have been infected.

   (4) I still don't see a lot of evidence that people are freaking out way too much. For example, I just went to a church where old people were still packed en masse and kissing icons. At least some of New York's bars were reportedly packed before they were closed. Boris Johnson shook everyone's hands in the Covid Ward. Rudy Gobert made sure to touch every microphone. People are protesting to open the economy back up. The draconian restrictions put into place have only modestly lowered the rate of infection. Even in Italy, six weeks of quarantine have only halved the daily rate of infections. To change people's behavior, you need them to "freak out" to certain degree.

   In any case, I am also opposed to a prolonged shutdown. Only, my alternative is Vo-style ubiquitous testing, or at least doing a vigorous "Test and Trace", plus some surveillance testing, ramping up N-95 mask production to wear in public, and other kinds of surveillance (cops with temperature helmets, like in China), etc. Nevertheless, it's clear now that Covid-19 poses a significantly higher danger in terms of mortality than the flu does.