I’ve been reading a lot of the contrarian (primarily from fellow righties) COVID19 opinions, and I wanted to work through them in good faith. I find most of them pretty unpersuasive. As they say, THREAD:

Obvious caveat: I am hardly an expert. I care about this stuff for its own sake and also because in my day job I manage (mostly) other people’s money, and many of them have asked me about my views on the crisis as it’s evolved. So here goes.

To organize, I’ll start by summarizing the big arguments, and then I’ll walk through each of them one-by-one, though many of these claims are related to one another:

1. COVID is not very deadly, with comparable (or even lower) mortality than the flu.
2. There’s a good chance COVID has been here for a while, and is already widespread in the population, though asymptomatic in most cases. https://www.ft.com/content/5ff6469a-6dd8-11ea-89df-41bea055720b

3. The COVID death count is way overstated, as many people are being counted as COVID deaths without proper documentation.
4. The economic damage we’re causing with COVID-containment lockdowns is a cure worse than the disease, so we should end the lockdowns soon.

“1. COVID is not very deadly, with comparable (or even lower) mortality than the flu.”
There’s obviously a ton of uncertainty on the overall mortality rate this early. But there are a few suggestive data points:

Germany and South Korea are often cited. For some time, both had mortality rates below 1 percent (seasonal flu is .1 percent). But as the disease has progressed (COVID takes a long time to kill people) they’ve both seen their mortality rise to 1.8 percent (last time I checked).

Iceland is another case study for the optimists. Their mortality is currently .3 percent. But they’ve had very few cases (serology testing suggests about .6 percent of the country is infected) and active containment. It's always tough to draw conclusions from a small sample size.

What’s more indicative of how this will progress? The many countries with thousands of cases and mortalities ranging from 2 to 10 percent? Or a single nation with 1,500 cases? Even if you do want to use Iceland as a lower bound, the mortality there is 3x the flu.

A suggestive data point I’ve seen on mortality is actually from an Italian town, where serological testing suggested 65 percent had COVID. https://it.finance.yahoo.com/notizie/coronavirus-castiglione-d-adda-il-065252105.html.

Importantly, not all were asymptomatic, and it’s a small sample size, but it does suggest saturation. And that town of 4600 saw 80 deaths. In other words, the cost of saturation in that town is that 1.7 percent of all—not confirmed infected, but all—residents died.

Evidence suggests that Italy is substantially underreporting deaths. So that 1.7 percent number is likely low. https://twitter.com/j_cd_t/status/1245739661449977867?s=12. This is not the flu.

The most common response to the above is that we are overstating deaths. I’ll address what when I get to argument 3, below. Before that, a related claim:

“2. There’s a good chance COVID has been here for a while, and is already widespread in the population, though asymptomatic in most cases.”

This was the most plausible optimistic scenario to me. Serious epidemiologists suggested it. But with more data, I no longer think this is likely. The best way to test this is by seeing how many people have COVID antibodies.

First, in Denmark, a place that’s been hit reasonably hard, one serological test measured 1,000 people in the Copenhagen metro area, and another 244 in Central Denmark. 2.7 percent tested positive in Copenhagen. 0 percent tested positive in Central Denmark.

Italy is a disaster, but one northern town reported only 15 percent antibody presence for COVID. The town mentioned above that had higher antibody presence also had sky high mortality. The story is pretty consistent: more COVID usually means noticeably more death.

In Iceland, estimates are that .6 percent of the population has COVID antibodies. In Telluride, a Colorado town, fewer than 1 percent tested had definitively positive results. Finally, genetic data can provide some insight into how long a disease has been around.

The phylogenetics we have suggest the virus sprang up in late November in Wuhan, and first moved to the United States in January. This is not a timeline consistent with widespread transmission by March. See here: https://www.medrxiv.org/content/10.1101/2020.04.02.20051417v1.full.pdf

The data is very incomplete. This preliminary view could be wrong. But so far, there's no evidence of widespread transmission. Note that this doesn’t mean there aren’t a lot of asymptomatic cases. There definitely are, just not so many that we’re anywhere close to broad exposure.

“3. The COVID death count is way overstated, as many people are being counted as COVID deaths without proper documentation.”

There’s a way in which this claim is sort of interesting as an academic argument: effectively, how do we track deaths in an accelerating pandemic, where it isn’t possible to test everyone who’s dying?

A lot of folks have picked up on CDC guidelines that don’t require a test to confirm a case of COVID. Others have focused on the “with/or” question—just because someone dies *with* COVID doesn’t mean that’s the cause of mortality. A few responses.

First, people are right to say the situation is fluid and it’s tough to record every death accurately. But importantly, 1) this is true for every disease (do you think every flu death has a confirmed flu test? No, just check the CDC guidelines) and 2) miscounting goes both ways.

There’s a lot of reason to think we’re undercounting deaths. First, it has happened in nearly every major European country. NYC public health officials have argued that where normally 20-25 people die per day at home, that number has been higher than 200 as the crisis has peaked.

But the truth is that we won’t know until we can look at the all cause mortality numbers, and see how they’ve spiked or not in response to COVID. (To be clear, I have some qualms with how useful the all cause numbers will be here, but they’re probably the best we’ll have.)

Many have argued that all cause mortality numbers are dropping lately (can’t find the tweets). This is wrong: CDC usually takes months to fully tabulate the data. There’s been a lot of misinformation here—a suggestion that pneumonia deaths are dropping because health officials

are overstating COVID numbers, for instance, is widespread on twitter. All of this stems from the fact that our mortality numbers don’t fully come in for months. One exception here is NYC, which usually reports vital statistics rapidly (and has done so for many years).

This chart from @lymanstoneky shows recent mortality stats, including this year (red line) and confirmed COVID deaths (black line). This is a *striking* image: nearly all of the increase in all-cause mortality comes from confirmed COVID deaths.

Finally, I think the “with/or” debate misunderstands something pretty fundamental about comorbidity. Most people who die of a respiratory illness (including, yes, the flu) aren’t totally healthy people.

If a dude with a heart condition gets the flu and dies, but would have lived another 10 months absent influenza, did he die of the flu? This shit is complicated, but that doesn’t mean there’s a conspiracy to inflate COVID deaths.

TLDR: all of the mortality estimates are imperfect right now. Health systems are doing their best to record the info accurately, we won’t know for sure what those numbers look like for weeks (at best), but there’s nothing unique about COVID in this respect.

“4. The economic damage we’re causing with COVID-containment lockdowns is a cure worse than the disease, and will lead to a lot of human suffering. So we should end the lockdowns soon.”

I’m sympathetic to this argument, in part because a lot of people freak out if you even mention the economic cost. This isn’t about dollar worship or stock prices—a lot of conservatives are genuinely worried about the anxiety, stress, and death that comes with a severe recession.

That said, one big problem with this argument is that it overstates how much of the “lockdown” has come from policymakers.

The Ohio governor closed restaurants at 9pm on March 15. Here’s the foot traffic data from OpenTable for Ohio, March 15 through March 5 (compared to the same days in 2019). This is not business as usual:

On March 17, Bloomberg ran an article about the UK’s approach to herd immunity, making it seem as if London was operating normally.

But that same OpenTable data suggested restaurant traffic was down on March 17 by 89 percent, and had been down substantially the weeks prior. Again, this was *before* the UK government closed restaurants and bars.

Some have pointed to Sweden’s laissez-faire approach to the virus as a model to emulate. Set to the side whether that’s accurate (it’s incomplete at best), Swedish economic forecasts are grim.

The most recent estimate I’ve seen from Swedbank is a 4 percent 2020 contraction. Goldman’s late March estimate for the US was a 3.8 percent contraction (they’ve since further adjusted down to 6.2 percent for the US).

In other words, a lot of the social distancing appears organic, and independent of the policy response. That doesn’t mean good policy can’t help (or bad policy hurt), but the idea that our economy just hums along absent lockdown orders from the president and various governors...

...is implausible. It turns out that people get freaked out about catching a deadly disease and adjust their behavior accordingly.

Further, China’s manufacturing indices suggests continued stagnation. India is locked down. Western Europe’s hospitals are overrun. Singapore appeared to have things under control, but has shut down again. Japan delayed the Olympics, and appears to be getting worse.

You can criticize globalization all you want (I do all the time). But a consequence of globalization is that you can’t thaw your own economy while the rest of the world is frozen, even if concerns over the virus are overblown. American politics can do only so much.

This doesn’t mean people don’t have valid gripes. The IHME model is and was a bad model for the present crisis, for reasons Andrew Gelman goes into here: https://statmodeling.stat.columbia.edu/2020/03/29/the-second-derivative-of-the-time-trend-on-the-log-scale/

Many experts have offered worst-case scenarios without qualifying them, thus eroding trust in public institutions when those scenarios fail to materialize.

But the failure of a given model—or more accurately, the failure of certain officials and media types to properly contextualize those models’ purposes and limitations—doesn’t mean that this crisis isn’t real.

One final thought on the modeling: much of it has been “curve fitting” meaning it assumes a pretty symmetrical bell curve. But I’m increasingly persuaded that the exponential decay side of the bell curve isn’t actually happening. See this data: http://nrg.cs.ucl.ac.uk/mjh/covid19/#rates-peaked.

My basic view is this: the virus is very deadly compared to the flu; it has not spread widely in our population; informal social distancing has done a lot to slow the spread (and more formal measures have helped too).

The problem with my view is that it suggests the toggle between “return to normal” and “viral spread” is very sensitive. If the average person reduces the number of people they come in contact with by 30 percent, you get lower spread.

If they reduce that number by 50 to 75 percent, even better. R0 doesn’t exist in a vacuum. But social distancing is *inherently* economically destructive. Even informally, even before any lockdowns.

This is the best paper that highlights how informal social distancing has likely already begun to flatten the curve: https://nber.org/papers/w26917 .

So what do we do? Good question, and I don’t know the answer (even if I’m right, and I may well be wrong). Continue to research good antivirals and vaccine candidates. Make everyone wear masks. Figure out how to make social distancing less economically destructive in the meantime

Stay safe everyone. This thread is already too long, but I’ll update where appropriate.

It was late and so I’m sure I made mistakes. I’ll try to correct them here. Someone pointed out that Iceland is doing PCR (not blood test for antibodies) testing. Given how rapidly they’re testing I think those results are still informative, but should have been more clear.

(Note to self: be more careful when using google to translate from Icelandic.) Also the @lymanstoneky chart above is for NY STATE. Not NYC.