Created 3/21/20. Latest update 3/21/20 8:45 PM. Formatting changes (which are coming) and minor editorial corrections will not be counted as updates. Suggestions for sources and for other questions welcome. At least for now, the answers here are predominantly my own opinions, downplaying other points of view.
1. Why are we so concerned about this virus, which so far (as of mid-March) has killed many fewer people than an ordinary flu?
2. How do differences in testing frequency and reporting practices affect reported spread rates and death rates?
3. What factors affect actual death rates?
4. What policies have worked best in reducing the spread rate?
5. How is government enforcement of social distancing justified by a difference between private costs and social costs?
6. How can a short, unpaid vacation for some sectors of the economy have such large ramifications?
7. What are the long-term economic consequences of this crisis?
8. Why aren’t policymakers taking into account the economic cost of closing restaurants and shutting down other forms of economic activity?
9. What about fiscal and monetary stimulus?
1. Why are we so concerned about this virus, which so far (as of mid-March) has killed many fewer people than an ordinary flu?
The key to the answer lies in the words “so far.” The virus seems to spread at a phenomenal rate, with cases doubling more than once a week. If the number of deaths were to double once a week, then starting from about 200 deaths on March 15, by the end of May the total would be 200,000 deaths, which is about ten times the number that we get from ordinary flu.
The Imperial College paper made an extrapolation that warned of the possibility hundreds of thousands of deaths if social distancing were not encouraged.
2. How do differences in testing frequency and reporting practices affect reported spread rates and death rates?
The reported spread rate depends on the actual spread rate and the trend in testing frequency. For example, in the United States, testing ramped up the week of March 16, and the reported spread rate rose above 50 percent per day on some days, but that is probably well above the actual spread rate. It is very difficult to estimate the actual spread rate as long as testing protocols are changing.
Death rates are also very unreliable. Some people have died of the disease without being tested for it, so that they do not count in the virus death statistics.
Deaths occur with a lag, so that if the number of cases is increasing, the reported death rate will understate the true death rate. To see this, assume that the true death rate is 1 percent, and death occurs 5 days after someone becomes a “case” by testing positively for the disease. Suppose that on day one there are 100 cases and on day five there are 1000 cases. As of day five, there will be one death (out of the original 100 cases), but dividing this by the number of current cases (1000), we would compute a death rate of 0.1 percent, which is far below the true number.
Other factors will cause the reported death rate to be overstated. Most important, many people who have the disease in mild forms will never be tested and thus will not count as cases. So the death rate relative to the number of reported cases will be high relative to the true death rate.
3. What factors affect actual death rates (not the reported death rates)?
An important factor is the quality of care. There was a relatively low death rate on the Diamond Princess cruise ship, and it is likely because the sick passengers and crew were well cared for. In Italy, the death rate from the virus was modest for a while, but it soared once the health care system became overwhelmed. The countries with the highest death rates appear to be those with the weakest health care systems and/or those whose systems are most overwhelmed by a large case load. But it is hard to be certain, given the challenges with reporting and calculating death rates accurately.
Age and existing health conditions are other important factors. Death rates for the very young are minimal, and death rates for the very old are high.
Although death rates among otherwise healthy people under age 60 and are low, they are not zero. Deaths of health care workers are particularly worrisome.
We should not treat the death rate as if it is some natural constant. When a large share of the cases is in the elderly population, as in the state of Washington in early March, the death rate will be high. When there are ample human and physical resources to treat patients, the death rate will be lower than otherwise. If new treatment protocols, including pharmaceuticals, prove to be effective, the death rate could plummet.
4. What policies have worked best in reducing the spread rates?
Policies to completely isolate those who have or might have the disease appear to have worked in China and South Korea. This seems to require strict quarantine and extensive surveillance. In South Korea, testing was widespread, as was “tracing and tracking.” When someone is sick, you try to trace back to see who might have made them sick, and you try to track down anyone else who might have come into contact with a sick individual. To do this, the authorities can use data on people’s phones to plot their movements. Presumably, this is not something that Americans are ready to allow.
If only sick people can spread the disease, then isolation of people with obvious symptoms can be effective. But if people who experience no symptoms or mild symptoms can spread the disease, then more general social distancing is required. Apparently, many experts believe that the disease can be spread by people who are not yet severely symptomatic.
5. How is government enforcement of social distancing justified by a difference between private costs and social costs?
The personal cost to me of risking infection is relatively low. Suppose I go to a crowded restaurant. Even if I get infected, I am unlikely to die, particularly if I get good treatment should my case be severe.
But the cumulative social cost of many people deciding to risk infection is high. First, they may infect others who have high risk. Second, as the number of people requiring treatment rises, the capacity of the health care system becomes strained, the quality of treatment falls, and the death rate rises.
Front-line health care workers are particularly inclined to stress the social cost of people risking infection. They see cases of young people requiring ventilators. They see fellow health care workers become sick and even die. They see the limits on supplies of equipment needed to treat patients.
6. How can a short, unpaid vacation for some sectors of the economy have such large ramifications?
Financial obligations don’t sleep. The small business that is temporarily closed and wants to keep its workers still has to make payroll. The individual who is temporarily laid off still has to pay rent and make car payments. The airline whose planes are grounded still has to pay the bank for the loans that it took out to purchase the planes.
Our society is highly levered. Many households have not saved enough to go even a few weeks without a paycheck. Many businesses make heavy use of debt financing and have little cash reserves. Banks deposits are supposed to be riskless and available on demand, even though the assets backing those deposits are risky, long-term loans.
When economic activity is running smoothly, this leverage is efficient. Households can enjoy goods sooner than otherwise by obtaining them on credit. Businesses can expand more by using borrowing than they could if they had to rely on cash flow from profits. Banks and other financial institutions allow households to buy homes, operate small businesses, and start new ventures.
But all this leverage increases fragility and interdependence. When households don’t have cash reserves, a short unpaid vacation means that they cannot pay rent. Then the owner of the apartment building cannot make the payment on its loans. Then the bank does not have the money to pay its depositors, unless it forecloses on the apartment building and sells it at a loss. Etc.
In a crisis, everyone tries to reduce their leverage. Households cut back on spending. Banks cut back on lending. Economic activity collapses.
7. What are the long-term economic consequences of this crisis?
Behavior will change, and economic activity will alter as a result. People who are considering going on a cruise or flying overseas will be more aware of the risk of contracting illness, and this will reduce demand. People involved in health care systems will be more aware of the risk of running into a surge in demand, and they will try to expand capacity.
People will become more facile at undertaking work, education, and conferences remotely. Businesses that help foster remote interaction will expand. Businesses that are built around large conferences and meetings will shrink.
As businesses develop supply chains, they may become less concerned with maximizing efficiency and a bit more concerned with ensuring robustness. They may prefer sources with more redundancy rather than sources that minimize cost.
An open question concerns what will happen with leverage. The financial crisis of 2008 caused some de-leveraging in the short term. But households and businesses soon got back to operating with high levels of debt and low margins of safety. Perhaps this time people will be more traumatized than they were in 2008. Perhaps this time government authorities will not be as sanguine about rising dependence on credit in the economy.
8. Why aren’t policymakers taking into account the economic cost of closing restaurants and shutting down other forms of economic activity?
This question implicitly assumes that the alternative of allowing all businesses to operate would be costless. That is, it assumes that we could avoid most of the cost of this crisis if we did not “panic.”
In fact, even if the government does nothing, many individuals and businesses will curtail their activities, based on what they know about the virus. Many private actors were ahead of the government in canceling travel and conferences. The economic costs were already going to be significant, even without government advice or mandates.
More important, an alternative strategy of allowing or encouraging ordinary economic activity could prove even more costly in the long run. One can imagine a scenario in which the health care system gets overwhelmed, causing many people to die who otherwise would not, both from the virus and from other treatable illnesses. There could emerge a social climate in which people fear and suspect strangers. This low-trust social climate would itself hinder ordinary economic transactions.
There is no way to avoid having the economy take a major hit from the crisis. Without a doubt, some of the policy responses will make things worse than would be the case otherwise, and others will make things better. It will be difficult to determine which is which, both in real time and in hindsight.
9. What about fiscal and monetary stimulus?
I am skeptical about both. Conventional monetary stimulus involves the Fed paying for financial instruments with money that it creates at the stroke of a pen. It’s a subtle way of manipulating rates of return in financial markets. That tool now seems to me beside the point, sort of like the steering mechanism on the Titanic after it already hit the iceberg.
Instead, I think we need a lender of last resort. I have suggested giving households and firms government-backed overdraft protection. Here is further elaboration.
Fiscal stimulus means the government taking money out of someone else’s pocket to give it to you. By someone else I mean a future taxpayer. The future taxpayer may face an explicit tax bill, or that taxpayer may end up paying via the “inflation tax:” giving up some real resource in exchange for government-issued money that is going to shrink in purchasing power.
Fiscal stimulus works because you need to pay your bills today. You don’t care about the future taxpayer. The politicians sure don’t.
As the recipient of the government payment, it looks like you are getting a gift, as opposed to the overdraft privilege or credit line. If you borrow using a credit line, you know you’re supposed to pay it back. But it doesn’t look like you have to pay back your gift. Instead, the person who pays back your gift is the future taxpayer. That future taxpayer could turn out to be you, but you won’t think about it that way.
The political lobbying for these gifts will be intense. I am cynical enough to believe that whatever you and I receive will be but a fig leaf to cover up more extravagant gifts for special-interest constituents. One of the big reasons I prefer the credit-line approach is that interest groups won’t fight as hard for money if they know they will have to pay it back.
Under #1, I would add that no one is vaccinated for this, which is why people worry about it more than the flu.
I don’t thinks that’s all that big a factor. I never worried about the flu before there were flu vaccinations.
Flu vaccines were developed in the 1940s, although they were more commonly available later. And occasionally there are shortages. But still, not clear on what you mean here. You were unconcerned in the early 1940s?
Maybe flu vaccines were developed in the 1940s, but they were not widespread until recently. It’s maybe 20 years ago that they started recommending that everyone get them and I started getting a flu shot each year. Both before then and since then, I have had cases of flu. They were unpleasant, but not scary. I can understand that older and unhealthier people might view it differently.
Gotcha.
But still, I think part of the worry about cov is that no one is vaccinated, so there’s no herd immunity to slow down the spread. The flu shot mostly protects you (imperfectly) and partly protects others by reducing the chance you will spread it. No such option here yet.
I guess one more way to reiterate my original point is a thought experiment:
If there had been a vaccine produced immediately when this thing broke in Wuhan, and administered to the same number of Americans as the flu shot, do you think anything would be shut down right now in the US? Even at ~50% efficacy, that would significantly slow the spread, and most would feel protected enough to lead normal daily lives, with maybe a little extra hand washing. People would still get sick and die, but people would have much more confidence in containment.
I think the main factor in the fear is simply that the coronavirus is much more deadly than the flu. Sure, it would help if there were a vaccine for coronavirus, even if it were only partially effective, because that would lessen the chances of catching it. But I still think that the vaccine is not that big a factor in the lack of fear of flu; the flu is just so much less deadly and we are all used to living with it.
Yes, I agree that the fatality rate could be the more important fear factor, although I suspect it’s a bit lower than current estimates due to lack of testing in mild and asymptomatic cases.
As I said, I would “add” lack of vaccine to his answer for #1, not replace the original. The relatively high fatality rate could fit there too.
Our society is highly levered.
That’s true, but it’s a financial-level explanation. Debts are legalities; in principle they could be voided. On a real-economy level, the problem is that we have just-in-time supply chains that mostly go through China. If we don’t keep the supply chains going, we’ll quickly run out of many things, notably medicines. That’s more fundamental; Congress can’t just wave that problem away.
The anecdotal evidence from Dr. Peter Attia and Dr. Peter Hotez yesterday is that NY hospitals are seeing many young and healthy residents requiring ICU/intubation care. This did not happen in the Lombardy region of Italy.
If allowed, the passive tracing would only be useful from the start of the collection moving forward. If the NSA has been secretly tracking phone GPS data then now is the time to use it. Software professionals with geospatial database experience are required. All the major SQL database systems (including SQLite) have very good geospatial index support.
Phone GPS location data is a GREAT strategy moving forward. For back-tracing I’d recommend using airline flight travel data for the starting point. The Canadian confirmed travel cases from the U.S. would serve as a small data set as a starting point. Either way, the Tracing problem now is one of scale. This is the largest Call Center operation ever and it has to scale-up within days.
This simple thought experiment should make clear that every effort starts with a Fermi Estimate and extreme tradeoffs required for each step forward. Don’t let perfect get in the way of the good.
Testing will soon be divided into 1. Active Virus Testing, and 2. Anti-body Testing for those Recovered. All tests so far have been of type 1 but type 2 tests should be rolling out soon.
Arnold,
Thank you for this helpful FAQ (and for your daily insights about the pandemic). A few notes:
Re: 2. Random testing, at suitable scale, at regular intervals, very probably would yield highly useful information for analysis, behavior, and policy. Incentives (e.g., payments) might suffice to achieve willingness to be randomly tested. However, if feasible incentives aren’t effective, then the question arises: Would coercion be justified to achieve random testing in the populace?
Re: 4. Coordination of policy across jurisdictions probably is crucial to durable effectiveness of any local policy. Robin Hanson wrote (March 16):
“We are all trying to decide if we will coordinate & pay huge costs to squash #COVID19 until a strong treatment arrives, or if we will flatten the curve but accept that most will be infected. We need to coordinate as 1/2 of places going one way & 1/2 going other just won’t work.”
Source: Robin Hanson, Twitter thread:
https://twitter.com/robinhanson/status/1239704352769609728
Re: 7. International patterns of specialization and trade very probably will change (but how much?) because political preferences will favor strict immigration control and domestic production of strategic resources. (You touch on this dynamic in a previous blogpost.)
RE: random tests. In yesterday’s Ontario ‘s COVID-19 Update, Dr. David Williams said that their Sentinal Surveillance system that targets hospital patients without symptoms has zero positive, 1821 negative, and 166 pending so far.
Even though the deaths counted toward coronavirus are unreliable and approximate, I find this graph of death rates per population useful and sobering. It compares coronavirus deaths per million population of the US and Italy on a log scale over time. Things to note:
– Both countries’ graphs are nearly straight lines, meaning both countries’ death rates are increasing exponentially.
– The two lines are very close to parallel, meaning the US is on a similar track to Italy. It looks like the US is 2.5 to 3 weeks behind Italy.
It will be worse. New York is Wuhan with mega-city-wide uncontrolled COVID-19 spread. The Lombardy region in Italy is 10 million people.
The spread can be stopped in a focused three week sliding window. Every day we need to plan the next 21 days until it is stopped.
And two new questions:
Is cost-benefit analysis of pandemic policy feasible and ethical? Sources:
https://scholar.google.com/scholar?hl=en&as_sdt=0%2C39&q=cost+benefit+social+distancing+life+years+saved&btnG=#d=gs_qabs&u=%23p%3DHa_whonqZicJ
Should relevant authorities (e.g., CDC and AMA?) proactively and transparently establish special pandemic triage criteria and mechanisms for allocation of scarce medical resources and care? If so, what should be the criteria and mechanisms?
Sources:
https://en.wikipedia.org/wiki/Triage
https://blogs.bmj.com/bmj/2020/03/09/covid-19-triage-in-a-pandemic-is-even-thornier-than-you-might-think/
https://www.statnews.com/2020/03/10/covid-19-gets-bad-are-we-prepared-for-triage/
There are two risk levels: getting the virus early, before widespread adaptations that improve the risks, and getting the virus after those adaptations are in place. The adaptations are likely to take at least until the second wave.
Getting the virus and surviving it makes you a near risk free worker, especially during the first phase. Assuming they don’t present a transmission risk, which is likely, this group will become a different class of citizen, with more value in the marketplace and more freedom.
I think answers to #5 and #8 need to be reconciled. If private business and individuals curtail activities on their own (#8), that suggests that many of the costs of infection actually are private, contra (#5). Indeed, individuals incur a private risk/cost of getting sick at a moment when the health care system is likely to be overburdened. Thus, expectations of future overburdening create private incentives for individuals to lesson their own risk of infection. Think of it this way: removing price ceilings not only incents additional supply, it also incents conservation on the demand side. An overburdened health system creates a sky-high price — treatment unavailable at any cost — which in turn creates private incentives to minimize the risk of needing such treatment.
The main gap between private and social benefits/costs is that high-risk individuals (seniors, others with health conditions) collectively benefit when low-risk individuals (young, healthy) avoid infection as that reduces high-risk individuals’ risk of infection. However, an environment with low-risk of infection is non-excludable and non-rivalrous. Hence, each high-risk individual would like all the other high-risk individuals to compensate the low-risk individuals to socially distance, then free-ride on the reduced risk. So, there is a case for government to force all high-risk individuals to compensate low-risk individuals to socially distance. But, as far as I can tell, there has been nothing in our social distancing policies that recognizes, let alone addresses, that the costs of social distancing are imposed, at best, uniformly on high-risk and low-risk alike, while the benefits accrue primarily to high-risk individuals. In fact, the costs might also be more heavily borne by low-risk individuals as many high-risk individuals, e.g., seniors, don’t go to work even during normal times. School closings also hurt low-risk young more than high-risk seniors. So, arguably, the long-term economic costs of social distancing are likely to be borne disproportionately by low-risk individuals even as the benefits accrue primarily to high risk individuals.
#7) One long-term consequence will be a test of Arnold’s Null Hypothesis on education. We will probably end up canceling school, or at least have tried replacing with online education, at all levels for the better part of a semester. It will be interesting to see in a few years whether we can detect any resulting difference in educational outcomes. I suspect that we won’t for subjects that are cumulative in nature like English and Social Studies. For sequential subjects like math, where one must master one topic as a prerequisite for moving to the next, we may find some effects. Talented students will be able to learn this semester’s lessons on their own and/or catch up next fall. I’m not sure that average students will be able to do so.
I also don’t think it will be politically feasible to expect such students to graduate one semester later. Instead, they will be socially promoted as if this semester was not cancelled, and the curriculum in the fall will be adjusted. They will graduate with one semester’s less education. That probably won’t matter for college education, if one views college as primarily signalling. It may make a difference, again, for K-12 education in sequential subjects like math. That would mean, of course, that college curricula might have to be adjusted for future students with one semester’s less preparation.
“That would mean, of course, that college curricula might have to be adjusted for future students with one semester’s less preparation.”
No “of course” about it. Won’t happen at all. There’s no “freshman year schedule” and hasn’t been for decades.
“So, arguably, the long-term economic costs of social distancing are likely to be borne disproportionately by low-risk individuals even as the benefits accrue primarily to high risk individuals.”
This is what utterly infuriates me, the automatic assumption that we’re doing the right thing. We almost certainly are not. We basically have a choice between the *risk* (not even the certainty) of overwhelming our hospitals in the short term and destroying our economy. This should be a no-brainer, and yet these idiots decide to destroy our economy.
It’s obscene how many people approve of this. I’m glad at least to see a contrarian view form, but everyone who’s shrieking for shut down just ignores them.
I think the contrarian view will become stronger if the social distancing needs to continue beyond 1-2 months. Just as it was wrong for people to treat coronavirus as “just like the flu”, it’s wrong for people to say that social distancing is “just like an extended Christmas vacation”. Social distancing for 2 weeks is a lot different from 2 months.
I think one of two things (or both) will happen:
1) exponential growth in confirmed cases continues for the next 1-2 weeks while testing catches up, even if social distancing is slowing down the true infection rate. This will lead people to claim that it isn’t working
2) eventually, the effects of social distancing (and possibly warmer weather) will slow down the growth. Those who were against shutting things down will claim that it was all overblown the whole time and it was going to slow down anyway.
“It’s obscene how many people approve of this”
Many of the commentariat are senior academics, public officials, executives, etc., age 60+ so, while not necessarily really elderly retirees, perhaps at somewhat elevated risk. Many of the young commentariat, say progressive woke journalists, are infatuated with the idea that Americans’ excessive commercialism needs to be tempered by Lenten-like rituals of abstinence and atonement. This could be an example of Bootleggers and Baptists.
A new source, Re: Macroeconomics of the pandemic (and the tricky concept of trade-offs. A new NBER paper:
https://tinyurl.com/uamyq32
The paper attempts to model various scenarios (containment, treatment, medical readiness, vaccine, etc.). Here (below) is the abstract, which reports “the benchmark scenario.” Estimates of other scenarios are in the paper. Of course, the model parameters lack the hard data (e.g., experimental data) that Dr. Kling seeks about contagion mechanisms:
“The Macroeconomics of Epidemics
Martin S. Eichenbaum, Sergio Rebeloy, and Mathias Trabandtz
March 20, 2020
Abstract
We extend the canonical epidemiology model to study the interaction between economic decisions and epidemics. Our model implies that people’s decision to cut back on consumption and work reduces the severity of the epidemic, as measured by total deaths. These decisions exacerbate the size of the recession caused by the epidemic. The competitive equilibrium is not socially optimal because infected people do not fully internalize the effect of their economic decisions on the spread of the virus. In our benchmark scenario, the optimal containment policy increases the severity of the recession but saves roughly 0.6 million lives in the U.S.”
The authors emphasize that the models don’t include some key factors (basically factors that Dr. Kling has highlighted in various blogposts):
“we note that our model abstracts from various forces that might affect the long-run performance of the economy. These forces include bankruptcy costs, hysteresis effects from unemployment, and the destruction of supply-side chains. It is important to embody these forces in macroeconomic models of epidemics and study their positive and normative implications.”
A new source, Re: Triage (a question I raised in a previous comment). An informative NYT article:
https://www.nytimes.com/2020/03/21/us/coronavirus-medical-rationing.html