Tuesday assorted links and non-links

1. Some mask production is being deregulated.

2. “Using country-, subnational-, and individual-level data, we show that Facebook has had a significant and sizable positive impact on citizen protests.”

3. Open source ventilators?

4. Automated contact tracing without giving up privacy?  And explanation.

5. What the SuperForecasters think.

6. A useful thread on mutation.

7. Scott Alexander on masks.

8. “People exhibit overconfidence in their ability to calculate exponential growth.

9. The Spanish flu rebellion against masks.  And how the Spanish flu shaped trust.  And 1919 Science piece on the lessons of the Spanish flu.

10. From my email: “One major issue for SMBs is that many owners are on the hook *personally* for lines of credit and other business loans and also for business credit cards. For example, my business has a line of credit that I had to personally guarantee and it is with the same major bank that holds my mortgage. Bankruptcy would probably mean I discharge the line of credit AND lose my house at the same time. I suspect other SMB owners are in similar situations.”

11. Aurlus Mabele, coronavirus victim (NYT) obit), RIP (music video here).  And Manu Dibango too.

12. “Based on the proposed methodological procedure, we estimated that the actual cumulative number of exposed cases in the total population in Lombardy on March 8 was of the order of 15 times the confirmed cumulative number of infected cases. According to this scenario, the DAY-ZERO for the outbreak in Lombardy was the 21st of January 2020. The effective per-day disease transmission rate for the period until March 8 was found to be 0.779 (90% CI: 0.777-0.781), while the “effective” per-day mortality rate was found to be 0.0173 (90% CI: 0.0154-0.0192). Based on these values, the basic reproduction rate R0 was found to be 4.04 (90% CI: 4.03-4.05). Importantly, by reducing the transmission rate by 90% on March 8 to reflect the suspension of almost all activities in Italy, we run the simulator to forecast the fade out of the epidemic. Simulations show that if the measures continue, the complete fade out of the outbreak in Lombardy is expected to occur by the end of May 2020.”  Paper here.

13. Why Germany is not doing as well as you think.  It is about the exponential function, yet again.

14. An overview on possible drugs and their status.


#5 "Superforcasters" are not so super. Totally wrong. We already see a slow-down of cases approaching in the US. Total US reported cases will be less than 100,000 by March 31 and in decline.

+1 and also great calculations with confidence intervals for R0 on #12...was waiting for someone to do this, as my intervals are not always so confident :(

Rnought is context dependent. What you want is rate of transmission under different scenarios, including the unconstrained (Italy) and the constrained (SKorea).

Yup, it's a classic example of an endogenous variable, rather than exogenous.

I'm going to remember that.

Am I misunderstanding that paper. It was written March 16th, so we have a week of data to validate their results by. In figure 5 they simulate the peak as occurring on March 10th or 11th, with less than 30k cases, right? Is there a reason why that's not as wrong as it seems to be?

they making predictions about 2021 not 2020

2021 is safe, and plenty of time to change predictions. This guy looks like an idiot with his prediction of only 500 American dead total. Richard A. Epstein, Hoover Institution, “Coronavirus Perspective,” published last week.

Not really. He was wrong but the data now seems like the number of dead in the US won't surpass 800.

Disinformation unless you show your sources.

Unfortunately the number will surpass 800 by tomorrow

That’s a high number, but don’t you really want to look at incidence in the context mortality rate? Isn’t that more a ratio of deaths:new cases as “mortality rate”, which has not changed significantly over time, even though new cases have risen.

It will probably take slightly more than a day, but yes the US number will easily go over 800 in the next 48 hours and over 1,000 in the next 96 hours and keeping climbing for at least a month.

New cases today are just a bit over half of what they were yesterday. I don't see the death toll climbing much more.

The US reports its numbers on a rolling basis unlike many other countries that only report once a day. Wait for the end of the day to see today’s numbers. I agree that this isn’t going to kill millions like some people are saying, but thousands is certain and tens of thousands are probable.

I think folks are freaking out about getting it and immediately fall into fatalism, when there’s adequate data that shows an overwhelming majority that gets it doesn’t die. Exponential data can *predict* but doesn’t guarantee. The virus won’t kill as much as the poor infrastructure of our healthcare sector and lack of funding for emergency preparedness.

Right, it’s also quite obvious that many other countries have avoided doomsday scenarios with either millions of deaths or economic depression, even China which was caught with its pants down is heading towards a not-terrible outcome. There’s no reason we can’t do the same.


Why would you say that's "quite obvious"? The only country you could make that case for at this point is China, right? That's the only place without community transmission isn't it?

Most countries in East Asia are doing well. Community transmission is low enough that individual cases can be identified, traced, and isolated without larger economic disruption.

If you observe recent developments I would say it's an open question whether they have merely delayed the large economic disruption, or avoided it. Singapore and Hong Kong have seen rising cases recently.

JWatts, see what I mean. No support. No data. No sources.


Prediction on a log scale is pretty weak sauce. At a minimum the buckets should overlap so we can see the difference between the 23 million and 230 million predictors.

Agree with “prediction on a log scale is pretty weak sauce”

With the responses pretty consistently falling with a normal distribution centered on the middle option, it sure seems as if there's a strong framing effect and they don't have a very good idea themselves. Responding under uncertainty using the answer options as a very important data source.

I think conflating reported/estimated is a terrible phrasing of the question. I think the number of cases will always be under reported. They are really two different numbers. just look at reported estimated deaths for swine flu (according to wiki at least)

18,036 (confirmed)
151,700 to 575,400 (estimated)

Also the ranges available for selection are huge . I think it would have been a better question to make one explicitly guess a number with no framing, and to explicitly choose estimated or reported.

-1 to their survey

One of the problems of governance where there is discretion is trust.

So, when Steve Mnuchin has the sole authority to hand out money and Donald claims he is the overseer, we have a problem.

One of the ways trust is regained is through having an oversight commission or approval process--a bipartisan appointed commission appointed and confirmed by the Senate. Another option is to expand IG's for the Agency and require direct reporting to Congress, and not just the Executive Branch. Or, it could be delegating oversight or approval to another agency like the SEC.

Too much power in one person if no immediate oversight.

What having a "commission" or "team" or "agency" does is devolve responsibility, which is the reason for having them in the first place. Mnuchin or whoever, if he's the agent, acquires the responsibility for the procedures to be followed, certainly with the input of others. Allowing that responsibility to be distributed means nobody gets the axe if there's a policy failure. Trump, for instance, is currently being blamed for any kind of failure at any government level, although he doesn't seem to be taking any heat for the Indian Health Agency keeping homosexual predator doctors on the payroll on reservations across the country.

They commission members look over each others shoulders and have a forum for expressing opinions. It's far from devolving responsibility. Commission or commission oversight is the last thing those who want to hide stuff in darkness want.

Looking at things that have occurred in the House over the past year or os would seem to indicate that these committees are very easily corrupted (see Adam Schiff) and used for nefarious purpose.

FYI - both Mabélé and Dibango appear to have contracted the disease and died in France...at the very least both died in Paris.

7. I appreciate your bringing SA posts to our attention. This was reassuringly thorough, as usual, and of particular interest to me in connection with a glance a few moments ago at Razib Khan's writeup yesterday, of a study presenting some evidence that humidity correlates with protection from the virus. Potentially encouraging news for the South? Maybe for a change, the first 20 minutes of the disaster movie won't contain the throwaway words "Houston is gone."

'Maybe for a change, the first 20 minutes of the disaster movie won't contain the throwaway words "Houston is gone."'

Is that a thing? I watch very few disaster movies so I wasn't aware of the line about Houston.

At least they're mentioning Houston. Smaller cities such as Portland don't get mentioned at all AFAICT.

I've heard it uttered several times, and I'm no connoisseur of disaster movies.

I think it filled a niche: big city that no one would miss. Hating on Houston was in vogue for far longer than, more recently, it's been held up as an exemplar of what freedom and the future look like.

Neither pole would ever be adopted by a Houstonian.

Houston is tarnished by Enron and the energy industry in general, which progressives love to hate on.

Oooh, let's not forget the satanic Astros!

Breathin' orange fire!

This has been bothering me for week: billions upon billions of dollars in budgets but no strategic stockpile of masks? Wtf? Are we that much of a banana republic??

Turns out there was one and we used it and never replenished it.

....The national stockpile used to be somewhat more robust. In 2006, Congress provided supplemental funds to add 104 million N95 masks and 52 million surgical masks in an effort to prepare for a flu pandemic. But after the H1N1 influenza outbreak in 2009, which triggered a nationwide shortage of masks and caused a 2- to 3-year backlog orders for the N95 variety, the stockpile distributed about three-quarters of its inventory and didn’t build back the supply.

It happened under the Obama administration, and therefore it hasn't been widely reported. You can guarantee that if it had happened under the Trump administration it would have been headline news.

It should be headline news. Trump had 4 years to do something about it but what's obvious is that public health isn't something he cares too much about. That and he doesn't listen to experts.

Obama had 7 years to do something about it and never did. So, it's a systemic failure not some kind of specific personal failure.

Italy has had fewer new cases for 2 straight days. Growth in new cases has been slowing steadily by about 1% per day since 3/7. If they can keep that up, it might fade out by mid April.

This preprint finds that cases are 15X higher than reported in France (which means it is a lot less deadly than we think): https://www.medrxiv.org/content/10.1101/2020.03.22.20040915v1

Germany seems to be doing well despite not social distancing until recently. 125 people have died from the virus and as with Italy, which is further along, has seen the percentage increase of cases decline:

Mar 17... 9,367... 29%... 27%
Mar 18... 12,327... 31%
Mar 19... 15,320... 24%... 28%
Mar 20... 19,848... 30%...
Mar 21... 22,364... 13%... 22%
Mar 22... 24,873... 11%
Mar 23... 29,056... 17%... 14%

0.4% mortality rate whereas Italy's mortality rate is 9.5% and 6,000 have died.

If Italy has under-reported cases by 15X, then morality is closer to 0.6%

The pre-print I linked for France estimates morality of 0.5%.

Then you have Germany, which is close to 0.5% without any adjustment for undetected cases.

The Italians are reporting 10x, not 15x. And Germany has been doing a fairly good job of isolating and tracking, making 10x a bit less likely, much less 15x.

I'm responding to the preprint that Tyler linked to: "Based on the proposed methodological procedure, we estimated that the actual cumulative number of exposed cases in the total population in Lombardy on March 8 was of the order of 15 times the confirmed cumulative number of infected cases"

#12: 3.3% of Lombardy's total population at that 15x level. Perhaps still a factor of 2-2.5x undercount of true infection rate, as I don't think even Germany is actually counting everything with much more representative samples.

At those levels, and considering the aged demographic who are hit and that their chance to die tends to be high anyway within a year to two years... well, OECD club probably still gains more life expectancy from lockdown, at least for a bit (and lockdown and rationing if anything is possibly good for affluenza).

But poor countries, even relatively well developed China, with younger demographics with fewer elders to protect, and where economic decline can be really important to health, might well easily lose more life expectancy, from lockdown than they gain from extended life expectancies for elders. Seems even entirely possible that China will in the short run lose more life expectancy from lockdown's economic effects than it gained from lockdown's effects in halting Cov19.

Unfortunately cases in Italy are higher today and deaths are 25% higher.

I see an 8% increase in cases over the day before:


8% increase in total number of cases today yes (I was referring to new cases today compared to new cases yesterday). Total deaths increased 12% today.

You said deaths in Italy are 25% higher then switched to 12% higher, the same as the day before but this keeps getting lower. It was a 19% increase three days ago.

Nobody in Germany thinks they are doing well, they are expecting a Italian, Spanish, or French style pandemic tidal wave to hit them in the next week or two. They are hoping, with a bit more time to prepare, a bit more time locked inside, and a bit more time to absorb lessons from Lombardy and the Elsass, that they will not completely drown like Italy or Spain.

Hope is a good thing, but realistically, it will be bad.

But the acceleration of German cases has been decreasing. Why would they expect an explosion, especially since they have finally started social distancing, which they didn't do much of until recently.

I've been bemused by the libertarian freak flag I see being waved here, but I will point out that the Congress's effort to saddle small business with an obligation to pay sick leave occasioned by Covid-19 is having an unintended but highly counter-productive effect: small businesses and medical practices are choosing to terminate their employees and shutter rather than face the prospect of trying to fund several months of sick pay to employees while revenues collapse. Sure, the law provides a tax credit against the employer's share of payroll tax, but the math doesn't work when revenues are collapsing. Unlike the sick leave law it replaces, the new law applies to any business with fewer than 500 employees. Of course, fewer than 500 includes 1. The law is effective April 2nd and businesses and medical practices are in a race to close up shop. Sayonara.

Nevertheless the sick leave provision is an utterly necessary one. These are not normal times and cannot by analyzed under "business as usual" rules.
Though at this point now the it is fast becoming moot since a great many people are having to close their doors entirely due to shut down orders, and this outs the employees in the unemployment line.

Companies under 50 can seek an exemption from the Labor Department if it endangers the business is my understanding.


As far as I know if you have a mortgage on a primary residence and declare bankruptcy you can affirm the mortgage and continue making timely payments and you will not lose your house-- it is a protected asset as long as you can meet the payments*. The same is true on auto loans.

* I think there is some high value limit on that, but it's high enough nit to snag most middle class houses.

It's state law- varies by state

It depends on which state you are in. In some, the house is potentially an asset that can be seized, in others it isn't.

Note, however, that the foreclosure process is extremely slow and very forgiving. That was one of the major impediments to recovery in the financial crisis. It took years to work through the backlog. So you will sometimes see advice that says in the event of hardship, stop paying your mortgage first. It'll hit your credit hard, but it will be a big help to your cash flow and it will take forever for them to do anything about your house. In the meantime you have the opportunity to find other solutions and for the economy to fix itself.

Oddly, as American's personal balance sheets have improved so much over the past few years, more have a lot of home equity and no longer have the walk-away option. I think something like 40% of households own their home outright. They don't face mortgages, though, which is nice.

#1 Does 3M control all of the inputs to production? I would assert they are not the only manufacturer of N95-equivalent masks on the planet. Other global manufacturers may be tapping the same vendors, who may or may not actually have capacity to create more. I assume SABIC, Saint Gobain, BASF, Dow have roles in the production of raw materials, are they ready for this? Perhaps. It's safe to say the C-suite replies "Yes, Mr. President." while the first-line is telling everyone to go home.

This is not a doomsday statement, I think we'll producing far over-capacity by the time this turns around. I'm just not sure this will be the set of events that actually turn it around.

4. Could this finally be a use for blockchain? Everyone has an anonymous ID, location information is stored on a blockchain so people could see where every anonymous ID has been and which are infected, without any central authority ever being able to track the anonymous IDs back to particular individuals?

sounds like it is still unclear whether/ how much covid19 is airborne
transmitted versus droplet?

Importantly, by reducing the transmission rate by 90% on March 8 to reflect the suspension of almost all activities in Italy, we run the simulator to forecast the fade out of the epidemic.
Arnold Kling would be suspicious.
I would say double check the coefficients, if they have statistical support them something good might be happening, not necessarily according to model.

Here's the daily growth in new cases in Italy each day since March 7. The slowdown in growth is real:

26.90% (March 7); 25.36%; 24.37%; 17.94%; 15.21%; 21.27%; 16.85%; 19.80%; 16.97%; 13.06%; 12.60%; 13.35%; 14.90%; 14.59%; 13.94%; 10.38%; 8.10%; 8.21% (today)

If that progress keeps up, new cases could be in the hundreds by mid-April.

Kevin Drum just wrote about Taiwan & Face Masks - to my delight:

Can We Get More Surgical Masks From Taiwan?

His numbers are slightly little out of date - new numbers just came in -Taiwanese mask production has risen from 70 Million/week mentioned there to 88.2 Million/week (12.6 Million/day).

#1...By deregulated you mean not having to buy liability insurance.

#6...How can you meaningfully predict mutations?

At some point, people with flu-like symptoms stop submitting to tests. Italy is probably reaching this point. The most remarkable thing about this pandemic is that the ratio of positives to negatives is fairly constant from the beginning the testing ramps to now. The only countries that really violate this trend are South Korea, Japan, and a handful of other Southeast Asian countries, but they may simply have reduced R0 much more quickly for long term cultural reasons- masks, for example, are not unusual items to see people wearing. It may simply be the case that R0 in those countries was never above 1 to begin with.

Before getting pessimistic or optimistic about numbers of new cases, just inspect the number of tests run and the percentage positive for a given geographic location. In the US, you can even look at the state level data. For example, I live in Tennessee which has recorded 615 cases, but they have will have run about 6,000 tests to get there (it was 3,000 when the case level was 280 a few days ago). I would expect that if you tested the 100,000 people in Tennessee with flu-like symptoms that is typical for a day in March, you would find about 10,000 cases COVID-19. It is just that in Tennessee, there isn't a general panic yet, so very few sick people bother to get tested. This will likely change over the next couple of weeks.

In other words, to identify a true peak, the ratio of positives to negatives in people with flu-like symptoms has to start decreasing. If it is just that the number of tests is decreasing, then you don't know much about the rate of spread.

I will point out that New York might be at this plateau of people getting the test- there were numerous stories the last few days about hours-long waits to get sampled- this will bottle-neck the number of people tested since it dismotivates getting a test. Yogi was right- it is so crowed no one goes there anymore.

5. Difficult to have confidence in predictions for the number of cases in the US one year from now when no one as many serious idea of how many cases of infection there are *now*.

They give 40% odds for a number of cases in the US between 2.3M and 23M in March 31, 2021. I (not a super-predictor but an hyper-predictor) would give a at least 10% chance that the number of cases today in the US is between 2.3M and 23M *today*.

After all, the only upper bound we have at this point
is that we have tested about 330,000 people, and got 43,000 positive,
so if the 330,000 tested were chosen randomly, it would give an infection
rate of 13%, or about 43 million people in the US !!

Of course, it is a gross underestimate because they people tested were tested because they were likely to be infected (because of symptoms or contact with other infected people). But my point is, we really don't know.
More than 2.3 million people infected now in the US is not at all absurd. If you think it is, what is your refutation?

Again, the only way to know this very important variable is to make a
systematic testing of a *random* cross-section of the population, say of 1,000 or 2,000 persons. This would give an estimate of the rate of infection. Then we could decide if fighting this pandemics is worth shutting down the economy for a week, for a month, for a year, forever, or not at all.

Agreed. Open geographic areas based on testing and implementation of programs that mitigate against spread.

While this pandemic is going on, people won't be going back to bars or restaurants in the same numbers they had before, so these sectors, including hotels, are toast for awhile, even if you open them.

"systematic testing of a *random* cross-section of the population, say of 1,000 or 2,000 persons. This would give an estimate of the rate of infection. Then we could decide if fighting this pandemics is worth shutting down the economy for a week, for a month, for a year, forever, or not at all."

I agree, it seems criminal this has not been done.

You kind of have to wonder if the CPC have done it in Wuhan, and then not told anyone, or else if they just don't have the capacity to do it at all... These seem like the only two possibilities at this point. They're certainly smart to think about doing it.

Or maybe many governments have done it, in secret, but have not published the result because they are frightening... But frightening it what sense? What would be very frightening is if we find a low number of
infected people, close to the ones which are announced daily. This would mean the infection is very deadly, and has a long time to run. Or it could be we find that there millions (or a dozen million) infected people in the US, which objectively would be reassuring, but maybe the governments are too afraid that it would terrify the population.

Well, I stop there, I am falling into the skin of complotist, and I don't like that.

Clearly the recent counter-narrative is driven by the belief that there are hundreds of thousands or millions of asymptomatic people out there, and therefore the mortality rate is much lower than believed. So people would be very prepared for that outcome and quickly dismiss the danger (probably appropriately). If the opposite, such theories would be pretty devastated. But, maybe they want the theory to stay alive to give people hope?

I'm more willing to believe that a secretive gov with an outbreak long in progress and a lot of capability to mobilize and keep results suppressed will have done it in secret than... Italy or the US, or places that haven't thought they had an outbreak at all. The likelihood that China or Russia have done it and kept it a state secret is pretty good, less so others. (Maybe there are ways they could do it in secret - divert some proportion of new blood donations etc.)

Re #13 "Why Germany is not doing as well as you think. It is about the exponential function, yet again.":

The author of the tweet Tyler is referring to here bases his analysis on a misleading chart. Or rather, the author is misjudging the chart as one making his case.

First: The exponentiality should to be looked at in terms of days passed since the first death was (or first X deaths were) reported in a country. Naturally, the longer day-0 date's back the higher the chances for exponential-growth to cede, due to policy actions.

Alas, of all the countries Eric Topol's tweet is showing in its screenshot, Germany was the latest to have a confirmed COVID-19 death [1]. E.g. France's first corona fatality happened 24 days before Germany's first. But the chart only goes back 13 days in time.

Second: Exponentiality is belittled by its base quantity. If Germany's death count continues to grow at the same rate of its neighbours its numbers of case fatalities will still be lower. Which can easily be noted as "doing as well as you think". Especially when looked in per-capita terms.

The chart highlights the doubling rate and minimizes the total numbers. Germany and Spain both are in the 3 days range, but Spain having 60% the population size of Germany, has ten times the number of corona deaths.

Finally: There are way too many variables included to draw conclusions at this point.

E.g. the portions of infected age-cohorts varies greatly across countries. People over 80 make up almost 20% if Italy's infected, while in Germany it's just 3%. That could still change in Germany and it's likely that this would make matters worse for Germany.

But that's still just an eventuality. And based on that analysis based on the wrong chart it still holds that Germany, as some other countries are, is doing rather well as we think.

[1] France: 2020-02-15, Iran: 2020-02-20, South Korea: 2020-02-21, Italy: 2020-02-23, USA: 2020-03-01, Spain: 2020-03-05, UK: 2020-03-06, Netherlands: 2020-03-07, Germany: 2020-03-10

#10. The risks of cross-default are why I always advise my clients *not* to put their business and personal accounts with the same bank. That said, the business's bankruptcy is still not a personal default unless the guarantor defaults on the guaranty. If your state permits you to have a "declaration of homestead" on your home, make sure you file one, which will protect a chunk of the equity from seizure. To quote Eugene Volokh, "this is not legal advice; if it had been legal advice, it would be followed by a bill."

Suggestion: Maybe some states could return to normalcy sooner than others. States that have the virus under control and aren't seeing many new cases could let businesses resume operations, say, with temperature checks at entrances. The federal government (CDC) could establish some guidelines under which states can allow businesses to reopen. Like number of new cases per day per capita or rate of decline in new cases or something. Percent contacts traced and isolated maybe. Kinda like the stress tests for banks done after the financial crisis. This would also incentivize states to get the virus under control since they would become, at least temporarily, magnets for economic activity. To prevent people from states with outbreaks from spreading the virus around, you could allow states to set up checkpoints to test for the virus. Temperature check, saliva swab, leave an address and phone number at which you can be reached. Of course this would require lots of tests to become available.


Fun Fact: Businesses are also on the hook for lost or stolen credit cards.

Limited liability only applies to personal credit cards, not business ones.

Comments for this post are closed