General update, April 12

1. Tyler Cowen offers some unsolicited advice from an economist to epidemiologists. More here.

I applaud epidemiologists for thinking exponentially rather than linearly. I boo them for going with computer models without seeming to appreciate how flawed data and parametric simplification make them unreliable.

I would add that the political process selects economists as advisers based on criteria that are uncorrelated, or perhaps negatively correlated, with wisdom. I could imagine the same thing happening when epidemiology mixes with politics.

2. Reader Aaron Lindsey has posted a 3DDRR spreadsheet that makes it easy to see how that indicator has behaved.

Do you want to understand how difficult it is to make epidemiological forecasts? Squint at the chart from April 1st on. Imagine a trend line from April 1 to April 6 extended to the present. The difference between that trend line and the actual behavior may seem small. But because of that difference, we may have twice as many deaths from the virus compared with what would have happened had the trend from April 1 through April 6 continued.

3. Another reader, John Alcorn, found this article.

We identified 103 possible work-related cases (14.9%) among a total of 690 local transmissions. The five occupation groups with the most cases were healthcare workers (HCWs) (22%), drivers and transport workers (18%), services and sales workers (18%), cleaning and domestic workers (9%) and public safety workers (7%). Possible work-related transmission played a substantial role in early outbreak (47.7% of early cases). Occupations at risk varied from early outbreak (predominantly services and sales workers, drivers, construction laborers, and religious professionals) to late outbreak (predominantly HCWs, drivers, cleaning and domestic workers, police officers, and religious professionals).

Their investigation looked at Asian countries other than China.

4. As economists start to think about how all this new government spending is going to be financed, you might want to re-read this post from 2010.

As of 1946, the ratio of debt to GDP was 108.67 percent. From 1947 to 1970, it fell to 27.96 percent. A substantial amount of the drop was due to the fact that the government ran a primary surplus in all but four of those years (the exceptions were 1953, 1959, 1962, and 1968), for a cumulative primary surplus of 43 percent of the 1946 debt.

I was surprised by this. I had not remembered these surpluses. One reason is that the primary surplus excludes interest payments. Including interest payments, the government mostly ran deficits, particularly in the 1960’s. Another reason may be that the U.S. only began to include Social Security surpluses in the overall Budget late in President Johnson’ second term. Had we used the “unified” budget from the beginning, the deficits would have seemed much smaller and we would have counted more surpluses.

Thanks in part to Social Security, which was running surpluses that were not included in the budgets as reported at the time, we were very fiscally responsible in the 1950s and 1960s, so we paid most of the World War II debt.

In contrast, today mainstream thinking is that “interest rates are low, so why bother worrying about the debt?” Olivier Blanchard, one of the leaders of the economics profession, takes this stance. I have never had much regard for him. Over the next decade, we’ll see who turns out to be right.

5. Tyler asks why Belgium is doing so poorly. I immediately looked for a sociological variable. It turns out that Antwerp and Brussels are two of the European cities with very high Muslim populations. The Netherlands, which has several cities that have large Muslim populations, also has a per capita death rate on the high side.

I am using Muslim population as a crude proxy for ghetto-ized living conditions, equivalent to Hispanics in New York City. I am not trying cast any ethnic aspersions on Muslims or Hispanics. What I picture is a lot of people who do not have the privilege of a social-distancing option. They cannot work in home offices or avoid living in crowded conditions. As the Wikipedia list says, “In Western Europe, Muslims generally live in major urban areas, often concentrated in poor neighborhoods of large cities.”

But Germany, which also has major cities with large Muslim populations, is doing relatively well. Berlin also has a high reliance on mass transit, which deepens the mystery of how Germany is containing the virus. This story credits early implementation of test, track and trace.

6. Some of you have expressed an interest in betting on the fate of colleges and universities. You will want to read this.

“Financially, many colleges have been struggling, facing a perfect storm which is going to be even more difficult now,” Robert Franek, editor-in-chief of the Princeton Review, told Campus Reform. “Their costs are up, but their tuition dollars are down as enrollments trend[s] down. The value of their endowments is also down, and with the economy in [a] downturn, alumni and corporate donations are likely to be less.”

The coronavirus relief bill signed by President Donald Trump on March 27 includes $14.3 billion for higher education, with $12.4 billion split between emergency grants to students and money to colleges “to address needs directly related to coronavirus” and to “defray expenses” from lost revenue, reimbursement, technology for distance learning, and payroll.

Tyler’s dire prediction for colleges is based on what he foresees as their inability to sustain the foreign enrollments that have been a crucial source of revenue. But note the huge bailout voted by Congress, even though it requires Republicans voting to give money to Democratic bastions. The higher education lobby has quietly become one of the most powerful and effective political operations in the United States.

The sociological variables

As we seek to understand different outcomes in different regions or countries, the temptation will be to look for explanations in terms of government policies. But I wonder how much of the differences were caused by sociological variables.

One variable might be ghetto-ization of immigrants. Some European countries have acquired large Muslim populations that have not been assimilated into the native culture. France, Sweden, and the UK come to mind, but perhaps there are others. Here in the U.S., New York has ghettos of Hispanic immigrants, and they account for a significant portion of deaths, perhaps in part because they are likely to work in occupations that require them to leave their homes. At the other end of the spectrum, Japan is notorious for not having a large immigrant population.

Another variable might be herding of the elderly. In the United States, we herd many of the elderly into nursing homes, and at least 10 percent of the deaths in this country have occurred among residents of such facilities. Perhaps living arrangements in Northern Italy produced an elderly herding effect. Japan has a very elderly population, but perhaps they live in more isolated conditions.

Are these or other sociological variables helpful in explaining the severity in Spain? What about the relatively low death rates in California and Texas?

PSST and the economic recovery

I keep forgetting that most people believe textbook macroeconomics instead of my preferred paradigm of Patterns of Sustainable Specialization and Trade. To understand how I think about the recovery from the virus crisis, forget everything you learned about “aggregate demand” or “aggregate supply” or anything you read in the newspapers, especially about all the things that the Fed is doing to “save” the economy. Instead, try to approach this with a fresh outlook, bringing no baggage with you.

I am going to make some points here that have the potential to offend people of all political stripes.

1. “We” didn’t kill the economy, if by “we” one is referring to the government. Most, if not all, of the actions that reduced economic activity were undertaken voluntarily by households and firms.

2. The economy is not going to “re-open” or “get back to normal” in response to government action.

3. The CARES Act and the Fed’s extraordinary interventions are more likely to hurt than to help the recovery.

Read on.

Continue reading

Antibody testing: overcoming Bayes’ theorem

Taal Levi writes,

Let’s talk about why it’s critical NOT to assume you are immune to covid-19 when you have a positive antibody test.

I will argue that this problem can perhaps be overcome by requiring three positive tests in order to prove immunity.

Let’s make Levi’s assumption that 1 percent of the population actually has antibodies. Having antibodies is a good thing, because it means you already have had the virus and we can hope that this makes you immune.

For rounder numbers, assume that the test gives a false positive reading in 5 percent of cases where people don’t really have antibodies, and also a false negative reading in 5 percent of the cases when people do have antibodies.

Out of 10,000 people, 1 percent will have antibodies. That is 100 people, with the other 9900 not having antibodies.

Of the 100 people who have antibodies, 5 will falsely be reported as not having them. 95 will correctly be reported as having them.

Of the 9900 who do not have antibodies, 495 will falsely be reported as having them.

Altogether, 495 + 95 = 580 people will be reported as having antibodies, but most of these people will not have them! So you would not want to tell all 580 people that they don’t have to worry about getting infected if they go out and play. Most of them in fact can get infected.

My recommended solution to this problem would be to require a second test for those who test positive the first time. The second test also has to be positive in order to say that the person has antibodies. Assuming that the results are independent, the chances of two tests incorrectly reporting positive result is .0025, or 25 out of 10000.

That still might be too high. But we can take the people who test positive on two tests and make them take a third test. Only if this last test is also positive would you give the person freedom to roam.

Note that if the probability of a false positive depends more on the person being tested than on pure chance, this proposed solution will not work.

3DDRR update

Today, the overall 3DDRR continued to edge down, to 1.40, and outside of New York it declined from 1.47 to 1.42

Recall the original post on the 3DDRR.

It would be nice to see, say, 1.5 by April 7, 1.3 by April 10, 1.1 by April 13, and 1.002 from April 16 on. That would raise April 4ths total deaths of 8314 to 17,834 (exactly, right?) on April 13

Well, we missed those goals by quite a bit, and as of April 11 deaths stood at 20339. As of April 6, we were doing nicely at 1.53, and it looked like we were on track to hit 1.50 by April 7. But that did not happen.

What I would like to see over the next week is a sharp drop in the New York 3DDRR without a big surge in other places.

Tyler Cowen at Princeton, annotated

Tyler Cowen talks about medium and long-term consequences of the virus crisis. I give it an A+.

For a couple of days I worked on a post on the economic outlook, which I scheduled to go up tomorrow. I think you will see a lot of similarity in our views. We differ in terms of tone. Tyler sounds detached and fatalistic. I sound cranky. Below are some more detailed comments of mine, sometimes amplifying his remarks and sometimes disputing them. Continue reading

General update, April 10

1. Ronald Bailey reports,

Over the last two weeks, German virologists tested nearly 80 percent of the population of Gangelt for antibodies that indicate whether they’d been infected by the coronavirus. Around 15 percent had been infected, allowing them to calculate a COVID-19 infection fatality rate of about 0.37 percent. The researchers also concluded that people who recover from the infection are immune to reinfection, at least for a while.

I cannot read German, and I am a little worried about how they calculated this. Suppose we find that as of today 10000 people have had the virus and 37 have died. I don’t think it would be correct to infer that the infection fatality rate is 3.7 percent. It could be that there are 50 people currently in the hospital, and if 40 of them subsequently die, then that would mean we should double our estimate of the infection fatality rate.

If we trust the 0.37 percent number (round it to 4 in 1000), we can work backwards from deaths to the number of people who have had the virus. So here is some speculative arithmetic. If 4 Germans have died for every 1000 who have had the virus, and the reported number of deaths is 2600, then the number of Germans who have the virus is 2600 * 1000/4 = 650,000.

We can play the same game for the U.S. Take 17,000 * 1000/4 and that gives 4.25 million people in the U.S. have had the virus.

2. Tim Skellet has some tweets about the German study. I noted this:

Overall in Germany: “So far, no transmission of the virus in supermarkets, restaurants or hairdressers has been proven”. #SARSCov2 was detected by PCR on “remote controls, washbasins, mobile phones, toilets or door handles” BUT NOT in infectious form.

. . .The entire #COVID19 outbreak in #Heinsberg district is traced back to a couple who attended a local Karneval festival. It all kicked off from there.

Pointer from Tyler Cowen.

3. I also liked another link from Tyler, the Dan Wang piece on life in Beijing during the lockdown there.

The virus and political alignment

Joseph C. Sternberg writes,

The oddity is that the left in most of the world has been so intensely critical of Sweden’s experiment. If this model works, it would hold out some hope that the coronavirus could be managed without putting millions of members of the left’s own blue-collar base out of work. Yet the prevailing attitude is less “let them try” and more “excommunicate the heretics.”

I prefer to use the three-axes model. For those of you new to this blog, the model says:

Conservatives like to frame issues in terms of civilization-barbarism, accusing their opponents of being on the side of barbarism.

Progressives like to frame issues in terms of oppressors-oppressed, accusing their opponents of being on the side of the oppressors.

Libertarians like to frame issues in terms of liberty-coercion, accusing their opponents of being on the side of (state) coercion.

For conservatives, the easiest way to frame this in civilization-barbarism terms is to cast China in the role of barbarians. President Trump has taken that approach.

Progressives instinctively reacted against this. Early in the crisis, the progressive framing, as articulated by WHO and some American progressives, was to charge that racism was behind the fears of the virus. They saw themselves as heroically fighting against anti-Chinese prejudice.

Since then, the progressive framing has become less clear to me. I have seen, but forgotten to bookmark, a few articles claiming that the virus crisis is harder on minorities because their death rates are higher and harder on women because they bear the burden of caring for children home from school. Those articles would represent oppressor-oppressed framing, but to be honest, I don’t see them as representative of what most progressives are saying at the moment.

For now, I see progressives as focused on claiming President Trump has badly mis-handled the crisis. It seems to me that they place a higher priority on that than on establishing an oppressor-oppressed narrative. Such a narrative may emerge later, perhaps in the report of the investigative commission that many progressives are calling for.

Libertarians are being driven bonkers. Myself included. I don’t have to repeat what I already have said. I see as villains all of those who seem to me to automatically praise activist government regardless of whether it helps while ignoring the possibility that the private sector can adapt effectively.

Of course, libertarians are backfooted by the undeniable fact that there are externalities here. If I behave recklessly, I can endanger others by infecting them or using scarce hospital resources.

Should it be legal to ride a motorcycle without a helmet or for a restaurant to have a smoking section? Many people would say “no.” Libertarians would be inclined to say “yes.” There is some of the same division over whether or not you should be allowed to eat in a restaurant these days. And libertarians are not winning the argument.

Questions from a commenter

The commenter asks,

I sometimes wonder, if it was 2007 (pre-iPhone) vs. 2020, would we have reacted the same way? Would it have been worse and for whom?

2020 seems to be predicated on two guiding principles:

1) the enduring belief in models to guide public policy regardless of their underlying ability to accurately predict actual outcomes to within any degree of accuracy

2) the always connected smartphone with never ending updates.

A few thoughts.

1. I think it is too early to say whether computer models helped, hurt, or made little difference. My bet now would be on “made little difference.” I think that most policy makers, including those in European countries, had to see with their own eyes what happened in Italy and New York before they would react. Asia was a different story, in part for cultural reasons but mainly because the SARS precedent made a more profound impression there.

2. I don’t think that the pre-iPhone reaction would have been the same. I take the view that our current technology blurs what used to be a distinction between our intimate world and our remote world. It used to be that our intimate world was our family, friends, and co-workers. We saw them in person. The remote world was celebrities, politicians, and people in the news. We saw them on television or in print.

Now, our intimate world and the remote world both show up on the same screen. We feel a compulsion to try to be celebrities in our intimate world (worrying about whether our friends “like” our posts on Facebook or Instagram) as well as a compulsion to become involved in the remote world.

During this crisis, I myself have torn down the barrier that I used to put up to prevent me from becoming too involved in the remote world. That is, I used to write a blog post and schedule it to appear two weeks later. Instead, I am writing blog posts to appear within hours, or even minutes after they are composed.

I think that the current media environment puts tremendous pressure on policy makers to appear to be doing something, and doing it quickly. Pre-2006, I don’t think that you get daily briefings led by the President.

The pressure on policy makers to react quickly may do more harm than good. Reacting quickly is not the same thing as making the best choices based on available information. In fact, it may be the opposite. In a pre-iPhone environment, maybe you don’t have Dr. Fauci making light of the problem by saying that young healthy people should be willing to go on cruise ships. Maybe WHO would have made its presence felt more on the basis of the epidemiology of its scientists than on the ideology of its upper echelons Maybe Congress would have been patient enough to come up with bill that more effectively addressed the problems of economic dislocation caused by the virus, rather than passing the worst legislation in history.

On the other hand, I think that the private sector was able to react more constructively than would have been the case 15 years ago. Individuals and firms knew enough to overcome bad regulations, either by ignoring them or shaming the government into dropping them. Weeks before government stepped in, many businesses canceled conferences, sports teams canceled games, and many of us adopted social distancing.

So I don’t have an overall evaluation of whether we are better off or worse off having our current media environment.

3DDRR update

Another slight uptick today, from 1.53 to 1.54

But the 3DDRR outside of New York went up even more, from 1.54 to 1.58

I expect the 3DDRR and the 3DDRRxNY to be lower tomorrow, perhaps even 1.45, just because the death numbers on Tuesday look suspiciously high, as if they borrowed numbers from either Monday or Wednesday. If you have a data series with that sort of daily noise, you may want to look at a longer time interval. If you look at a 5DDRRxNY, it is very close to 2, meaning that the death rate is doubling every five days instead of every three. The good news is that the trend for the 5DDRRxNY is down, but the bad news is that it is not falling fast enough to warrant saying that the worst is behind us.

Speaking of noise in the daily data, a commenter from Indiana pointed out to me the contrast between deaths reported by the covidtracking site that I use for the 3DDRR calculation and the Indiana Department of Health site.

Here is how I understand what is happening. Suppose that 15 deaths occur on Monday, but the Department of Health finds out about 5 of them on Monday, 7 on Tuesday, and the other 3 on Wednesday. As of Monday, it will have incremented its total cases by 5, which is what the covidtracking site will pick up as the incremental deaths for Monday. Then on Tuesday, the Department of Health will revise its Monday total up by 7, but the covidtracking site will pick up these 7 deaths as incremental for Tuesday. And on Wednesday, the Department of Health will revise its Monday numbers up by 3, but the covidtracking site will pick those up as incremental for Wednesday.

In a steady state, daily deaths and the proportions reported staying constant, this would not affect the covidtracking site. But if deaths are increasing (decreasing), that site will show the increase (decrease) with a lag. If something similar holds true for other states, then this will add noise to the 3DDRR calculation.