1. The WSJ reports,

Public-health experts generally agree that to reopen society safely, communities need widespread testing so officials can be confident that the number of coronavirus cases in the population is low and people who are positive can be quarantined.

The article is still talking about doorknob effects and how long virus can live on surfaces. Meanwhile, nobody has looked for evidence that people get the virus from surfaces.

Consider two policy objectives. One is to protect the health of people who are vulnerable, primarily the elderly in nursing homes. The other is to protect otherwise less vulnerable people from getting something worse than a mild case of the disease.

We do not have the scientific evidence to determine what is necessary or sufficient to achieve either objective. But “public-health experts generally agree” that testing is the answer. I would feel better about public-health experts if they generally agreed that it would be a good idea to run experiments and carefully evaluate hypotheses before pronouncing them.

2. Javiero is back.

First I want to focus on the Wuhan Commercial Vehicles Show (CCVS) that was held between November 1 and November 4 in the Wuhan International Expo Center. Besides Chinese manufacturers present at the show, including of course Dongfeng Trucks, foreign manufacturers present at the show included Mercedes-Benz, Scania, SAIC-IVECO, Michelin, and Palfinger.

Read the whole thing. I like his detective work. As he points out, his analysis implies that the virus was in Wuhan by early November.

One can think of government as playing two roles in the virus crisis. One role is to exercise control, meaning giving orders. The other role is providing information, including reliable data and analysis. My criticism of government can be summarized by saying that it has been too eager to use control, while in the area of information it has been derelict and incompetent.

Here is what I would do if I were in charge of the CDC.

1. I would have one unit focused on providing consistent, accurate information about deaths. Deaths would be reported by date of death. Deaths would be reported in categories: deaths with no relationship to the virus; deaths of people with the virus but caused primarily by pre-existing conditions; deaths that were caused by a combination of pre-existing conditions and the virus; deaths that were caused primarily by the virus. The CDC reporting unit would give clear guidance to health care workers on how to do this classification. Trends would be reported by age and by institutional status (nursing homes, prisons) as well as by geographic area.

2. I would have another unit charged with determining the prevalence of the virus. As you know, there are two types of tests, one for whether someone currently has the virus and another for someone has the antibodies to the virus. For each of the two types of test, the testing unit would use the testing procedures with the highest reliability, including re-testing people if that reduces classification errors. It would use stratified random sampling.

3. I would abandon all models that work with a single spread rate or a single infection fatality rate. Instead, I would work with the Avalon Hill metaphor and have a unit evaluate hypotheses relative to that metaphor. Some of these hypotheses can be tested using healthy volunteers willing to expose themselves to possible infection. Others can best be evaluated by studying cases of infection events and deaths. The idea is to better predict what happens in an encounter between an infected person and a person at risk of becoming infected.

This unit of the CDC would focus on how both the probability of infection and the severity of disease are affected by the following factors:

4. Characteristics of the person at risk for becoming infected. age; and pre-existing conditions, including obesity. (This cannot be tested experimentally, but the cases that we have seen could be evaluated more closely.)

5. Extent of symptoms of the infected person.

6. Type of contact between the infected person and the person at risk.

7. Duration of contact between the infected person and the person at risk.

8. Distance between the infected person and the person at risk.

9. Masks. Neither person uses a mask; only the person at risk uses a mask; only the infected person uses a mask. Both use masks.

10. Whether contact takes place indoors or outdoors.

My first choice would be for government to provide information on these factors and let individuals and businesses make decisions based on this information. My second choice would be for government to obtain this information and issue orders to citizens based on this information. The current state of affairs is that government issues orders without this information. As I see it, exercising control without information is the least desirable role for government.

1. Thomas Meunier writes,

While new medical treatments proposed to cure COVID-19 cases are required to be validated through controlled double blind studies, the benefits and risks of social distancing strategies are not subject to any comparative tests

we show that the available data exhibit no evidence for any effects of the full lockdown policies applied in Italy, Spain, France and United Kingdom in the time evolution of the COVID-19 epidemic. Using a phenomenological approach, we compare the evolution of the epidemic before and after the full lockdown measures are expected to produce visible results. Our approach … is focused on incident rather than cumulative data, and it compares pre-lockdown and post-lockdown trends. However, here, no positive changes are noticed in the trend of the daily death growth rate, doubling time, or reproduction number, weeks after lockdown policies should have impacts.

That is certainly the case with U.S. data on death rates.

Lockdowns have achieved a theatrical purpose. They allow politicians to pose as powerful leaders implementing a cure for the virus.

Pointer from John Alcorn.

2. A study by R.E. Field and others of a cohort of 500 COVID-19 patients at a London hospital. Many interesting findings, including

When the outcome of the ventilated patients is viewed by age and gender … no female over the age of 60 has yet left the intensive care unit alive and nomale over the age of 50 has left the intensive care unit alive.

Another pointer from John Alcorn.

1.12 overall, 1.16 outside New York. I think I will wait until Tuesday, May 5 to do another 3DDRR update.

1. Jeremy Samuel Faust writes,

there are little data to support the CDC’s assumption that the number of people who die of flu each year is on average six times greater than the number of flu deaths that are actually confirmed. In fact, in the fine print, the CDC’s flu numbers also include pneumonia deaths.

His point is that deaths from the flu in past years are greatly overstated. This reinforces the intuition that the novel coronavirus is worse than the flu.

2. Culture of looting watch, from the WSJ

housing activists in at least 15 cities, including New York and Chicago, are organizing rent strikes. They are calling on tenants to withhold May payments in hopes of provoking federal and state lawmakers to provide more financial support for renters.

3. Ricardo Reis looks to the period right after World War II as a precedent for how fiscal and monetary policy were entangled and then unentangled.

There comes a time when the central bank says that it should be focused on controlling inflation and no longer just on trying to keep rates low just to finance the debt. The finance ministry, of course, does not like that because now its job is going to be much, much harder. Potentially it’s going to have to deal with high interest rates, especially if it does not behave responsibly with respect to its deficits. As a result, this conflict arises. The Accord was, if you like, the peace treaty.

Pointer from Greg Mankiw. The interview with Reis is broad, recommended.

The accord to which he refers is a 1951 agreement that allowed the Fed to set a more independent course, raising interest rates. That was easier to do then, because the government had stopped deficit spending when the war ended.

4. A commenter pointed to a post by editors of The New Atlantis about the extreme differences across states in death rates. I’m not necessarily with them on the analysis, but I pass along the link because the map charts are eye candy.

We have a moral revulsion to doing controlled experiments on small groups of people. Yet we have no problem conducting uncontrolled experiments at large scale.

Consider the California governor’s order to close public parks and beaches. We are unlikely to learn anything from this experiment.

A controlled experiment would start with two sets of healthy young people. Some would stand outside in a setting where there is an infected person. Others would stand outside where there is not an infected person. Still others would stand inside where there is an infected person. And others would stand inside where there is not an infected person. We would compare outcomes.

Many people would find this idea repulsive. Yet we would learn something from it.

Years ago, I noticed the same thing about education. I happened to be seated at dinner next to a high-ranking career official in the Department of Education. I asked why educators rarely used controlled experiments to determine the efficacy of alternative methods. He said, “Would you want your child to be the subject of an experiment?” My jaw dropped. “They do it all the time!” was my reply. “They just don’t do it rigorously/”

Schools were often introducing new variations in curriculum. Like the decision to close the beaches, these were arbitrary changes based on intuition, with no plan to learn whether or not they were effective. Tampering, as W. Edwards Deming would call it.

Markets allow society to test the behavior of entrepreneurs. The social mechanisms for testing the behavior of politicians do not work nearly as well.

1. David Benatia and others write,

We found a nationwide 1.9 percent infection rate in early April, which is similar to the estimated prevalence in Austria (1.1%), Denmark (1.1%), and the United Kingdom (2.7%) as of March 28. Meanwhile, Germany’s 0.7% infection rate would rank in the lowest tercile of prevalence among U.S. states. The highest rates of infection in New York (8.5%), New Jersey (7.6%), and Louisiana (6.7%) are still lower than the estimated rates in Italy (9.8%) and Spain (15%).

Pointer from John Alcorn.

As I understand it, their approach is to look at how the change in reported cases per capita changes as you change the number of tests per capita over a short time period. That makes sense to me. The

Their results imply a total number infected as of early April of about 6 million, which is lower than the outlandish figures some have proposed but much higher than the reported cases at that time.

2. This story reports,

The number of nursing homes publicly reporting cases of covid-19 has doubled in the past week, with more than 1 in 6 facilities nationwide now acknowledging infections among residents or staff, a Washington Post analysis of state and federal data found.

…In five states – Maryland, Massachusetts, Connecticut, Georgia and New Jersey – the virus has struck a majority of nursing homes, the data shows. In New Jersey, second only to New York in total number of confirmed coronavirus cases, health officials have reported infections at 80 percent of the state’s homes.

Doubling the number of nursing homes hit in the last week would lead me to expect a dramatic increase in deaths a few weeks from now. I am cynical enough to predict that the media will attribute these deaths to “re-opening the economy.”

I really wish that deaths were consistently reported by age category, rather than as aggregate figures. My point is not to suggest that deaths of the elderly do not matter. My point is that our prevention strategy ought to be aligned with where the risks lie, rather than talking about “the” infection fatality rate.

3. From Nature,

parts of the world that do not have a policy of universal BCG vaccination, such as Italy and the USA, have experienced higher mortality associated with COVID-19 than places with long-standing universal BCG vaccination policies, such as South Korea and Japan

This has also been proposed as an explanation for why death rates are lower in Eastern Europe than in Western Europe.

1.

To date, the total number of reported patients treated with HCQ, with or without zinc and the widely used antibiotic azithromycin, is 2,333, writes AAPS, in observational data from China, France, South Korea, Algeria, and the U.S. Of these, 2,137 or 91.6 percent improved clinically. There were 63 deaths, all but 11 in a single retrospective report from the Veterans Administration where the patients were severely ill.

Source here. I recommend the whole article (short).

They are not talking about a controlled study. Remdesivir shows some small mortality benefits in a controlled study, plus it is not tainted by association with President Trump. Which drug would you want if you were diagnosed positive?

2. Concerning Sweden,

This individual-based modelling project predicts that with the current mitigation approach
approximately 96,000 deaths (95% CI 52,000 to 183,000) can be expected before 1 July, 2020.

Source here. That model-based forecast was made on April 10. As of April 29, one widely-used site gives Sweden’s cumulative deaths as 2462. The model predicts a peak in the first half of May. New deaths on April 29 in Sweden were 107. If this is the peak, we will not make it to the bottom of the 95 percent confidence interval for the model.

Pointer from Dan Klein. This story makes Sweden look good and a model look bad. So you know I am bound to like it. Discount accordingly.

A reader asks whether the excess deaths we are seeing now are mostly people who would have died later in the year, so that for 2020 as a whole the epidemic will not stand out. My thoughts.

1. For what it’s worth, the models say otherwise. Experts who look at models expect another wave of deaths as social distancing is relaxed.

2. I am willing to speculate against the models. What I am seeing is a very gradual decline in the death rate. The death rate will remain high enough over the next month to ensure that 2020 will be a bad year for excess deaths. But we may very well see months later in the year in which “excess deaths” are negative, because of the death in March-May of people who would have died later in the year.

3. During the first two weeks of April, we saw cumulative deaths rising at 10 percent per day or more nearly everywhere in Western Europe and in the most hard-hit U.S. states. Now, we are seeing increases of 10 percent per day almost nowhere. Not in Sweden. Not in Florida. Not in Texas. There is a very broad based slowdown. Too broad based to be accounted for by lockdowns, which were imposed in different degrees and at different dates across all of these regions.

a. Perhaps the slowdown is an illusion, and we are going to get a second wave. That is the story that the models want to tell.

b. Perhaps the virus has already attacked the weakest victims. Recall the Avalon Hill model, in which there are offensive factors and defensive factors. Perhaps the people with weak defensive factors have already been attacked, and what remains are harder targets: elderly in well-run nursing homes that are keeping the virus out; or people who themselves have a decent defense against dying of the virus.

c. Perhaps the offensive factor is down because the virus has mutated to be less deadly. I remember an early Weinstein-Heying podcast where Bret said, “The virus doesn’t want to kill you. It wants to get into the future.” Their point was that the virus would increase its survival chances by mutating into a less deadly form. They did not give a time frame for this benign mutation trend to emerge, and it may be improbable that it has taken place already.

By the way, here is a long NYT article on excess deaths.