The video of my debate with Justin Wolfers is out. Normally, I note highlights but I won’t do so here because this has been a busy week and it hasn’t slowed down even on Friday afternoon.
I think I’ll give a better perspective after I’ve had time to watch the whole video this weekend and make notes on it.
The first minute is Gene Epstein giving ground rules, 17 minutes of my case, about 15 or 16 minutes of Justin’s case, about 5 minutes of my rebuttal, about 5 minutes of Justin’s rebuttal, then about 25 minutes of discussion among Gene, Justin, and me, and then about 5 minutes each of summing up (me first and then Justin.)
On one technical issue that divided us–how to measure the loss in GDP in a quarter–I promised that if I were right, I would post on it and if Justin were right, I would post on it. Justin, Gene Epstein, and I have been emailing about it this morning, and I now think the probability is greater than 0.9 that Justin is right and I was wrong. But it’s an interesting technical point, and I think I’ve come up with a clean way of laying it out. That might not happen until Sunday or early in the next week.
A mutual economist friend of Justin and mine, who I didn’t know until this morning is a friend of Justin’s, wrote me the following and gave me permission to quote:
I was talking to Justin Wolfers about something else and he mentioned how he found you genuine and a great debating partner.
The feeling was and is mutual.
READER COMMENTS
Arnold Kling
Apr 24 2020 at 5:48pm
My post-mortem will be up on my blog tomorrow. After Wolfers made his first point, I had to shut it off. The notion that keeping the lockdown going can save a million lives is one that I am very confident is false, but it is also non-falsifiable. The debate ended at that point, as far as I am concerned.
David Henderson
Apr 24 2020 at 6:20pm
Interesting. I did think the 1 million number was hugely exaggerated and it is true that to make his case he needed a number that high or almost that high.
Alan Goldhammer
Apr 25 2020 at 10:20am
While I hope Dr. Kling is correct, I’m less certain. Looking at the 2010 census data shows the population of Americans over the age of 70 to be just over 40M. That age group is most at risk of mortality from SARS-CoV-2. In the absence of a vaccine or prophylactic drug treatment, one only needs a CRF of 2.5% for 1 million deaths and this is only one population cohort (and of course it is the one most at risk based on multinational observational data). My own opinion is that with modest reopenings and continued public health measures coupled with testing, we will end up with a CRF of o.3%; this is the CRF that Britain observed for the 1957-58 flu epidemic. If I have done the math correctly this is still 900,000 deaths [300,000,000 people x 0.003; I’ve rounded down the US population to the 2010 number] assuming everyone in the US gets infected and there is normal disease progression. This would be the upper boundary limit. We don’t know at what point herd immunity kicks in and assuming it is at a lower % of the population than other viral diseases, a high mortality result will still be observed.
I’m finding the modest prevalence of strokes in younger SARS-CoV-2 patients disturbing.
Don’t get me wrong, I am a strong believer in getting things back up and running. The problem remains that this virus is subject to the Rumsfeld Doctrine.
John Alcorn
Apr 25 2020 at 11:19am
In a remarkable study, Maryam Farboodi (MIT), Gregor Jarosch (Princeton), and Robert Shimer (Chicago) explain that a standard epidemiological model (SIRD) ignores spontaneous, private social distancing (adaptation by individuals and firms). They sketch and critique a naïve model, with facts in hand:
In other words, the authors estimate that the number of deaths in the US would be 600,000 if no one adapts behaviorally and if the pandemic burns through in one peak. If the model is credible, then Justin Wolfers’ estimate of 1,000,000 deaths is hyper-naïve.
In other words, What Arnold Kling said.
John Alcorn
Apr 25 2020 at 2:28pm
To be fair, Justin Wolfers (during the rebuttal portion) concedes David Henderson’s point, that people and organizations adapt by voluntary social distancing and self-closures of ‘non-essential’ economic activities. Then Dr. Wolfers argues that simple extrapolation from NYC’s fatality rate predicts half a million deaths in the US. Picking up on Dr. Henderson’s statement that voluntary adaptation accounts for 90% of economic contraction, Dr. Wolfers argues that lockdown causes only marginal economic harm. What he neglects, I think, is that lockdown delays economic adjustment (reallocation), erodes liberty, increases political strife, and creates contentious, thorny issues around how and when government should lift lockdown.
If I understand correctly, Prof. Wolfers rests his case on two points:
(a) The precautionary principle: “If the worst of the epidemiological models are right, then hundreds of hundreds of thousands, potentially even millions of people may end up dying as a result of this. So the balance of risks here is incredibly asymmetric. You only need to put a small probability on me being right, to think that lockdowns make sense from a public policy perspective.” (Cue time 00:32:14)
(b) A strategy of radical suppression (rather than mitigation): “I’m not going to be able to go to a public park, a shopping center, a school, without confronting risks—unless we can beat the bug, unless we can eliminate it; and that’s what the lockdowns are all about.”
Listeners will judge whether protracted suppression—until effective treatments or a vaccine might be developed—is feasible in the USA; i.e., whether the USA can emulate Taiwan.
If suppression is the plan, then I hope that summer will greatly reduce transmission rates and severity of illness, and buy time to develop test-trace-isolate, treatments, vaccines, and any other remedies …
John Alcorn
Apr 25 2020 at 2:38pm
Cue time of Dr. Wolfers’ insistence on suppression (“elimination”) is 00:44:44.
David Henderson
Apr 25 2020 at 5:35pm
John,
Thanks. I’m virtually positive that I didn’t say that “voluntary adaptation accounts for 90% of economic contraction.” Justin suggested that I had said that, but I made very clear that I hadn’t. In fact, if I had thought that voluntary adaptation accounts for 90% of economic contraction, I would not have been bothered debating because 10% is a small number.
John Alcorn
Apr 25 2020 at 8:31pm
David,
Thanks for catching that! I checked. Right you are!
At cue time 00:02:50 you state, “A 30% drop from that is $1.6 trillion. If only half of that drop is due to the lockdown—and I think probably more than half was—that’s $800 billion.”
As far as I can tell, Dr. Wolfers introduces the 90% figure, but as a measure of how many people practice social distancing. Then he uses the 90% figure also as a measure of the cost of the lockdown—and proceeds as if you agree. Here is what Dr. Wolfers says:
“The final thing [David Henderson] says, and I think this is absolutely critical, which is we need to distinguish between the cost of a shutdown—a lockdown—versus social distancing. And he’s right. What does a lockdown do? Well, before the lockdown, 90% of us were social distancing. So what the lockdown does, is it makes 10% more people socially distant. These are the people who didn’t care much about their own lives before. Presumably, they didn’t care much about others. These may well be the super-spreaders. These may be the people we really really need to cut down on. But even if that’s not true, if even without the government lockdown, we do 90% of this distancing, 90% of us stay at home […] the benefits of the lockdown are only 10% as large, but so are the costs of the lockdown. So David’s point—I can concede it entirely, it’s an important one—what it says is that this entire debate about whether to have lockdowns, if in the absence of a lockdown we’d have a 90% ‘lockdown’, then it’s like all he’s saying is that the present debate that he and I are having is only 10% as important as he and I used to think it was. But if he’s going to use this 10% to knock down the benefits of the lockdown, he does need to use the 10% to knock down the costs in equal measure.” (Segment 00:43:25-00:44:38)
I should have checked carefully his construal of your position against what you actually said!
AW
Apr 25 2020 at 4:37pm
Regarding lockdowns, don’t we have several natural experiments, like Sweden? If the stakes are really a million lives or western lockdowns, we can review the data afterwards to see how different countries performed. Far from unfalsifiable!
Mark Bahner
Apr 27 2020 at 3:16pm
Yes, Arnold Kling is right that there’s no “second Earth” with a “second USA”…but I don’t think makes the assertion that 1+ million people “would have” died “unfalsifiable.”
There are not only countries like Sweden, there are even different state levels of lockdown, and different state timings of lockdowns. There is also the possibility of, for example, examining how many deaths were associated with long-term care faciliities, prisons, and meat processing plants. The lockdowns can’t be credited with saving lives at those locations. For example, let’s suppose fully 100 percent of the deaths in the U.S. were to come from long-term care facilities, prisons, and meat-processing plants. Then the deaths would be the same, with or without lockdowns.
So I’m with you, there are plenty of ways to have a pretty confident judgment about whether U.S. COVID-19 deaths would have been over 1 million without lockdowns.
Tyler Wells
Apr 24 2020 at 9:23pm
Yes, he needs to have a case not just that the coronavirus will take a million American lives, which seems fantastical, but that the difference in lives between lockdown or no lockdown is a million. Eventually, that has to be some evidence or we will be back at the same place every 5 to 10 years with every new pandemic.
Michael
Apr 24 2020 at 9:33pm
Well, the evidence would suggest that what we are experiencing now is far rarer event than once every 5 to 10 years.
Tyler Wells
Apr 25 2020 at 10:23am
There is a pandemic at least every 10 years, plus major influenza outbreaks. If the response is “lockdown until we know something (or have a vaccine)” and given how far into coronavirus we are and how little we still know, we will have many more lockdowns going forward.
https://www.cdc.gov/flu/pandemic-resources/basics/past-pandemics.html
Michael
Apr 25 2020 at 12:00pm
Yes, there is. But this particular one is biologically quite different than anything in recent (ie, past few decades) experience.
Put another way, there are reasons, related to the biology of the infectious agent, why we are locked down now and were not locked down for SARS, H1N1, etc.
Jerry Brown
Apr 24 2020 at 10:06pm
I just want to thank you for doing this debate and allowing me to view it.
So- Thanks! It was an hour and a half well spent for me. You didn’t change my priors much though. I started off more on what ‘Wolfers side’ was and ended up there as well. But the discussion about what if masks were worn in public could it be loosened up a whole lot did make an impression, and I know that is something Arnold Kling has been advocating for a while, and it does seem to make sense even if, rather than protecting the wearer from infection, it worked more to protect others from the wearer.
Well, Thanks.
David Henderson
Apr 25 2020 at 12:03am
You’re welcome, Jerry. If the mask idea makes sense, though, I’m not clear on why I failed to persuade you. Masks are not the same as lockdowns.
Jerry Brown
Apr 25 2020 at 3:55am
Professor, it is just the lack of data and that we are only four months into this thing, which is part of the reason for that lack. But I can’t but feel that our government has failed us here also. Seems they wasted about two months warning time to me at least.
But if we had some data that masking yourself cut down the ‘externalities’ sufficiently, then I think I am all for it. If we had some masks that we could buy easily instead of trying to make our own- well that would be even better. But that is a different problem.
I don’t know if you saw the antibody survey NY State did and announced yesterday. I did my own very amateur calculations based on that and the results were horrific to me. I will support the ‘lockdowns’ in my state for a few more weeks assuming we also are getting data and providing time to develop treatments.
David Mohler
Apr 26 2020 at 12:05am
I also did amateur calculations based on the NY antibody survey, as well as the LA County survey, but came to a much different conclusion; it’s the first confirmation that the fatality rates are substantially lower than originally thought – under 1%.
Jerry Brown
Apr 26 2020 at 1:06am
David Mohler, yes my calculation was below 1% also. I believe I came up with 0.76% case fatality rate overall using the 13.9% already infected with an estimate of 21,000 deaths in NY State. I later saw that Cuomo was using 15,500 deaths, which he admitted was low, and projected a 0.5% case fatality rate. Hopefully, both of us are very wrong, as in over estimating this.
Either way, if projected to the entire US population these numbers are right in Wolfers range depending on what percentage of the population actually gets the virus. Potential 1 million dead seems worth a pause to try to do something about it to me.
David Mohler
Apr 26 2020 at 1:19pm
Given NYC’s population density and a public transit system ideally suited for rapid dissemination of a virus, I’m skeptical that the rest of the country will see NYC rates. LA County’s rate, again by my amateur calculations, is less than a third of NYC. Time will tell…
Craig Yirush
Apr 25 2020 at 12:18am
David,
Have you addressed what I take to be Justin’s strongest point – that voluntary social distancing had already begun to reduce economic output, maybe by as much as 90%, before the lockdown, which really reduces the share of the economic hit attributable to the lockdown. And also means that if we opened up tomorrow, there’s no reason to think that consumption will come roaring back.
Craig Yirush
Apr 25 2020 at 12:47am
Edit – I see that David did address it in his closing. I spoke too soon.
Michael
Apr 25 2020 at 7:28am
A few thoughts, now that I have had a chance to listen.
1. I agree with David that this debate was exceptionally well-handled by all involved. This was great to listen to, probably as good as one of these things can get.
2. I think during both David’s and Justin’s presentations, my focus was on the uncertainties (I very much agree with David’s use of that term instead of ‘risks’), which I think are massive and interlinked. Off the top of my head, relevant uncertainties here include characteristics/behavior of the virus, the benefits and costs of social distancing (and other voluntary measures), and the additional benefits and costs of lockdowns over and above those of social distancing alone.
3. Also, it may be worth thinking of school closures as a step between social distancing and lockdowns – a government closing public schools is different from the government saying ‘you must shelter in place’ – that comes with its own set of costs and benefits. Key points: it is plausible that children readily contract and pass on this virus while not necessarily experiencing symptoms, meaning that schools could be a key vector of spread; also, school closures (along with the uncertainties around safety) have a disporoportionate impact on working parents with school-aged children.
4. I find Justin’s 1 million to be a plausible estimate of where we would be if we had just decided to let the virus run wild, and I agree with his point that we need to be thinking about the full range of potential outcomes, not just the most likely ones. We’ve already seen at least 50,000 deaths in under 2 months, the bulk of those coming within 1 month, and it is continuing. That is with some very aggressive measures (social distancing, school closures, and lockdowns) that have reduced spread.
John Alcorn
Apr 25 2020 at 9:29am
A study by Nicholas Banholzer et al., “The estimated impact of non-pharmaceutical interventions on documented cases of COVID-19: A cross-country analysis,” finds that school closures and shelter-in-place orders (lockdown) do little to reduce the number of new cases: “the mean effect size for school closures is among the lowest across the policy measures considered in this study.[…] lockdowns have only a small added value.” (p. 14)
See Figure 2 at the link.
By contrast, closures of venues and of “non-essential” businesses substantially reduce new cases. However, as David Henderson pointed out in the debate, venues (e.g., sports arenas) and firms also voluntarily and quickly close until they can implement credible, prudent protocols. Venues and firms reckon with reputation, and with the fact that customers and workers develop strong preferences for social distancing.
Border closures and bans on large gatherings also have substantial effects. These are government bans.
Of course, these are preliminary findings. Don’t hang your hat on one study. I keep a watchful eye on Sweden.
Arnold Kling reminds us to distinguish two rationales for any bans to reduce news cases:
(1) To flatten the peak of the first wave, in order to remain within critical medical care capacity in the short term. This shifts the curve (delays contagion) and buys time to increase hospital capacity and to implement more effective measures to shield eldercare facilities and persons with risky pre-existing health conditions.
(2) To suppress prevalence, and then to prevent new spread, until effective treatments or a vaccine will be discovered. Prevention of new spread requires an effective test-trace-isolate regime. This is the Singapore or Taiwan model.
It is quite possible to fall between two stools here.
robc
Apr 26 2020 at 8:13am
Ban on large gatherings mostly happened voluntarily…NCAA and pro sports leagues shut down before the government bans.
Mark Bahner
Apr 25 2020 at 12:30pm
But look where the deaths are occurring!
Deaths in long term care facilities
How have school closures and lockdowns reduced spread and deaths in long-term care facilities? School closures and lockdowns haven’t helped reduce spread and deaths in long-term care facilities, have they?
And look at the deaths by state. Between 40-45 percent of all deaths in the U.S. have come from only three states: NY, NJ, and CT. And the majority of the deaths in those states are probably related to New York City. So what is the sense of lockdowns in states that have not demonstrated anywhere near the potential for COVID-19 death that those three states have demonstrated? In fact, why even have lockdowns in, for example, northeastern and southwestern NY state?
John Alcorn
Apr 25 2020 at 9:51am
I should clarify that the costs of falling between two stools would be (a) deeper, longer economic contraction, (b) erosion of liberty, and (c) greater political strife—whilst delaying, but not greatly reducing cumulative deaths.
Michael
Apr 25 2020 at 12:02pm
I am very concerned that we will fall between 2 stools as you put it.
Dylan
Apr 25 2020 at 10:42am
As I mentioned in the last post, I found the debate informative and learned things from both sides. One area that I thought was underexplored was the idea that government mandated closures, supported with programs to support temporarily unemployed workers (like the PPP) could have a benefit both in terms of saving lives, and in reducing the long-term economic impact from the disease compared to voluntary measures alone.
As both Justin and David acknowledged, much of the social distancing and shut-downs were happening before the official lockdowns went in place. I know many bars and restaurants had started shutting down here in Brooklyn, as had the $0.99 store where we buy most cleaning products, corner grocery stores, the beverage wholesaler, etc…However, the places that were still open were busier than ever. I met a friend at a bar just a couple of days before the official lockdown, and there was a sign at the front door telling people to maintain safe distances, but the place was shoulder to shoulder standing room only (and this is a big bar that normally has maybe a third as many people on a Friday night)
A number of big businesses that had people that could work from home, were nevertheless not allowing it, because they thought people were more productive when they came into the office. I think David even mentioned in a post a few weeks back knowing people that had quit their job because they weren’t allowed to work from home?
My point is, that it seems at least plausible that a voluntary and disorderly shutdown could relate in as much or more economic damage, while still allowing the virus to spread fairly effectively. There was a hope that a ordered shutdown, with government providing insurance, could allow the shutdown to be shorter and more organized, and let people get back to work more quickly and with less friction.
It’s far from obvious that is what is actually happening, but markets are behaving as if that is at least a decent probability.
Mark Bahner
Apr 25 2020 at 11:58am
The methods of “shelter-in-place” or shutting down “non-essential businesses” doesn’t make sense in terms of “buying time” to “implement more effective measures to shield eldercare facilities”.
“More effective measures to shield eldercare facilities” can be implemented immediately. And “shelter-in-place” or shutting down “non-essential businesses” don’t help eldercare facilities at all.
Here are measures that can help shield eldercare facilities that can be implemented immediately, and don’t have anything to do with “shelter-in-place” or shutting down “non-essential businesses”:
A) Rather than require all people to stay at home, simply don’t allow visitors to eldercare facilities for several months. Note that this is a very extreme measure, but probably results in significant protection. The measures that follow
B) Scrupulously monitor eldercare facility workers for COVID-19 symptoms or possible asymptomatic infection, e.g. multiple daily temperature measurements, monitoring for coughs, daily certification that none of their family members have COVID-19, etc.
C) Provide personal protection equipment (e.g. washable masks and washable gloves) for eldercare workers.
D) Provide portable HEPA (high efficiency particulate air) filters to provide HEPA-filtered air (which has more than 99.97% of particles 0.3 micrometers in size removed) blown (gently) into the face of patients whenever they have visitors or are being assisted by eldercare workers.
E) Scrupulously monitor eldercare patients for symptoms of COVID-19 (e.g. multiple daily temperature checks, monitoring for coughs).
F) Etc. There are many more measures. Literally hundreds.
To summarize and repeat:
“Shelter-in-place” and closing “non-essential” businesses do nothing to protect eldercare patients, and there are measures that can be implemented immediately to protect eldercare patients. (So there is no need for any delay to provide time to implement these measures.)
Mark Bahner
Apr 25 2020 at 12:02pm
Oops. Measure “A” should have said (italic to emphasize the missing words):
A) Rather than require all people to stay at home, simply don’t allow visitors to eldercare facilities for several months. Note that this is a very extreme measure, but probably results in significant protection. The measures that follow are probably preferable to this measure.
John Alcorn
Apr 25 2020 at 12:15pm
Yes, the preliminary evidence indicates that shelter-in-place doesn’t reduce new cases. Moreover, private adaptation by individuals and firms (social distancing, restructuring of workplace shifts and spaces, self-closure of venues, etc) goes a long way. Therefore, the crucial early policies by governments should focus tightly on shielding eldercare centers and vulnerable demographics (and, I would add, stratified random testing to ascertain and monitor prevalence). One caveat: Sweden’s public-health authority explained that bans on gatherings of more than 50 persons, and enforceable rules about social distancing in restaurants, cafés, and so on, are are necessary (in Sweden) in order to flatten the peak and thereby to buy time to increase hospital capacity.
Mark Bahner
Apr 25 2020 at 12:49pm
But even if shelter-in-place did reduce new cases, it would not necessarily help long-term care facilities, because the average person simply does not visit long-term care facilities. (Or the average person could simply be banned from visiting long-term care facilities.) So the issue of COVID-19 at long-term care facilities is totally separate from the issue of COVID-19 in the general population.
Note: I wish I’d thought about this issue, and commented on it, during David’s debate with Justin Wolfers!
P.S. Good man, Justin Wolfers, but his incorrect thinking on several key issues has definitely led him to the wrong side in this debate. 😉
Phil H
Apr 25 2020 at 12:57pm
I’ve only watched the opening arguments so far, but I thought they were good. In particular, I thought DH presented his argument extremely well, and laid out the terms of the debate very clearly.
The first part was very America-centric, in terms of not taking much evidence from other countries. That was a shame.
For the most part, I just think that DH’s bright line distinction between government action and private action is not supportable. If the virus is to be beaten, it will take massive, coordinated action. Government is just the name of the mechanisms through which people coordinate on the largest scale. The idea that nationwide coordination could/should somehow go around or in parallel to the very institutions that exist to coordinate the country (i.e. the government) is just bizarre and redundant.
So I couldn’t get onboard with DH’s side in the debate. But it is interesting, thanks to both.
Tom West
Apr 25 2020 at 3:58pm
One aspect about the general truism that businesses want to stay in business and thus are unlikely to take steps with respect to safety that endanger their long-term reputation is that mostly holds true when there is a long-term.
I’ve had some experience with people living in a condo created by a respectable builder with an excellent reputation that hit financial difficulties during construction. In a desperate (and ultimately futile) attempt to survive, they took every legal shortcut, and a few illegal ones as well. Something they had never done in the previous 20 years.
Perfectly understandable. Reputation doesn’t matter if you lose everything.
I suspect there are going to be a large number of businesses that are on (or even beyond) the edge of going bankrupt, whose set of incentives for taking the costly steps that might ensure customer safety are going to be sorely tested.
(And yet knowing this, if things open up, I’ll end up prizing culture over reason and visiting restaurants, trusting they wouldn’t put my health at risk.)
Bill Conerly
Apr 25 2020 at 3:59pm
I wish David had hammered more on the knowledge problem. Individuals make choices with intimate knowledge of their circumstances. Government orders ignore most details.
Banning recreational travel sounds like a great idea, but twice I traveled to visit a dying parent. Those two trips are precious to me. So maybe a government edict okays travel to see a sick parent; but what about an uncle who was like a father to someone? The government cannot know the details of a person’s situation.
In the business world, I could go into the office, (I’m a consultant to some essential businesses) but I’m working from home voluntarily. Across the street from my office is a recreational products company. Suppose an engineer goes into the office, alone, because access to equipment is needed to finish a new design. The government edict says this is not essential and treats it as harshly a nightclub reopening, even though it’s a pretty safe activity.
Most people are smart enough to figure the trade off between safety and achieving other goals. The governors and mayors could certainly have tried asking for voluntary adjustments first.
Otherwise, David and Justin both did a great job, though I miss the occasional humorous cheap shot.
John Alcorn
Apr 25 2020 at 6:14pm
I apologize for scattering my various comments here and there across this thread. I would like to return to a crux of Dr. Wolfers’ case, the precautionary principle: “If the worst of the epidemiological models are right, then hundreds of hundreds of thousands, potentially even millions of people may end up dying as a result of this. So the balance of risks here is incredibly asymmetric. You only need to put a small probability on me being right, to think that lockdowns make sense from a public policy perspective.” (Cue time 00:32:14)
A question arises: Is it obvious that the cost of a small risk of 1,000,000 deaths, adjusted for expected longevity, is greater than the cost of a small risk of a Great Depression? Dr. Wolfers compared the risk of 1,000,000 deaths to the annualized loss of output from a quarter of a year under lockdown—not to the risk of a Great Depression. Where would consistent application of the precautionary principle land?
And, what if, say, a Great Depression increases risk of nuclear war?
IMO, David Henderson is wise to emphasize sheer uncertainty around the pandemic.
An ancillary comment: Dr. Wolfers categorically stated, “Deaths are not going to go up.” (Cue time 01:01:05). He is referring to deaths other than direct Covid-19 fatalities; and he is making an inference from death rates during previous economic recessions. (Although suicides increase during recessions, deaths from other causes decrease even more.) However, already there is preliminary evidence of substantial net increase in other deaths during the pandemic. See Kieran F. Docherty et al., “Deaths from Covid-19: Who are the forgotten victims?”:
https://www.medrxiv.org/content/10.1101/2020.04.21.20073114v1.full.pdf
Summary of findings:
“With the spreading global pandemic of coronavirus disease 2019 (Covid-19) there has been disruption to normal clinical activity in response to the increased demand on health services. There are reports of a reduction in non Covid-19 emergency presentations. Consequentially, there are concerns that deaths from non-Covid-19 causes could increase. We examined recent reported population-based mortality rates, compared with expected rates, and compared any excess in deaths with the number of deaths attributed to Covid-19. […]
There was an increase in observed, compared with expected, mortality in Scotland (+27%), England and Wales (+35%), the Netherlands (+60%) and New York state (+26%). Of these deaths, only 43% in Scotland and England and Wales, 49% in the Netherlands and 30% in New York state were attributed to Covid-19 leaving a number of excess deaths not attributed to Covid-19. […] A substantial proportion of excess deaths observed during the current COVID-19 pandemic are not attributed to COVID-19. This may indicate an increase in non-COVID-19 deaths due to changes in routine health care delivery during this pandemic.”
Michael
Apr 26 2020 at 11:00am
I think this is a problematic line of thinking, in the sense that I’m sure it would be quite easy for economists to model a tradeoff between “1 million lives” and “a Great Depression” but that isn’t a choice that is available for us to make, and such a model would be a poor proxy for the question Justin and David debated (should lockdowns end immediately?).
Major sources of economic damage during this crisis include:
Covid-19 itself. As Justin noted, deaths are economically costly. Caring for Covid-19 patients is a significant demand on resources. There are a lot of reports of sequelae of Covid-19 (such as organ damage) that are both a human and economic cost.
The voluntary measures people and businesses are taking to minimize risk. Now is a terrible time to be in the bar/restaurant/hotel/airplane/sports/movie/live entertainment business. Since restuarants can be a very high fixed cost/low margin business, the percentage of people opting for voluntary social distancing measures may not need to be that high to disrupt businesses.
Lockdowns. Obviously costly, but the lockdowns could end tomorrow and there would still be major, economically costsly disruptions.
There’s nothing good about Covid-19 itself, but the other measures have both benefits and costs. Voluntary social distancing mitigates some of the economic damge done by Covid-19 (most obviously the economic value of lives lost/saved) while doing its own economic damage. Same for lockdowns. There’s no simple tradeoff of lives vs economic damage available.
Viewed in isolation, the various mitigation strategies have been very economically damaging, that is for sure. But how much of that is damage the virus would have done anyway? Unclear.
I don’t think Justin meant to imply that all-cause mortality would decline during Covid-19, even as Covid-19 and suicidal deaths rose. We are certainly going to see (and are already seeing) a huge spike in all-cause mortality. But given the inadequate testing regimes in most countries, a lot of Covid-19 deaths are not being attribued as such.
Mark Bahner
Apr 25 2020 at 6:32pm
Quote of Justin Wolfers from the debate:
This quote doesn’t make any sense to me. Suppose every single restriction was lifted tomorrow, and every person made the personal choice to behave exactly as they had in December 2019? Presumably, COVID-19 deaths would increase. But if COVID-19 deaths increased substantially, why couldn’t all the restrictions simply be put back in place? And if the restrictions were put back in place, wouldn’t the deaths then slow, plateau, and eventually begin declining (as they currently are)?
How is it even remotely possible to potentially reach “even millions” of COVID-19 deaths in the U.S., even if all restrictions in all states and localities are completely lifted tomorrow?
Phil H
Apr 25 2020 at 9:57pm
“if COVID-19 deaths increased substantially, why couldn’t all the restrictions simply be put back in place?”
It’s not OK to experiment with human lives like this.
Mark Z
Apr 26 2020 at 12:32am
Doesn’t all public policy experiments with human lives though? We implement what we think is the best policy even though it is drought with uncertainty and if it fails we change course. It is not that choosing one policy of the two is the default and the other is a dangerous experiment (or whether it is is precisely what is being debated). The point is, the policy course is reversible, and to some extent the success or failure will be observable so we shouldn’t assume we will have to commit to the failed policy forever.
Mark Bahner
Apr 26 2020 at 8:03am
You don’t think rendering unemployed what might be more than thirty percent of the workforce (over 40 million workers), greater than the level of the Great Depression, isn’t an “experiment with human lives”? Specifically, what about young working families, just out of college or even high school, who basically have no savings? Isn’t this shutdown an “experiment” with how they will manage?
And who can even say what the U.S. federal deficit will be this year? Three trillion dollars? Four trillion dollars? Isn’t it an “experiment with human lives” about how younger generations are going to pay off that additional debt?
And how about the states? Isn’t it possible that some will be bankrupted by the expenses, absent tax revenues? Isn’t it an “experiment with human lives” to see how people in those states will manage? Or how about the people in other states who will take on not only the additional personal debt and federal debt, but even possibly the debt of states in which they don’t even live? Isn’t that an “experiment with human lives”?
P.S. You must be able to look far enough ahead to realize that the shutdown is eventually going to have to be substantially loosened. How is that not also an “experiment with human lives”?
Phil H
Apr 26 2020 at 11:40am
“Doesn’t all public policy experiments with human lives though?”
No. That’s like saying, we take risks just walking out of our front door, so we might as well walk down the middle of a busy freeway. There is a monster eating people right in front of us. The option of throwing more people to it is not a good one.
“Great Depression”
Don’t be hysterical.
“shutdown an “experiment” with how they will manage?”
Yes. Managing is one thing. Dying is another. Managing is better.
I live in China. We locked down for three months, and the epidemic ended. America should do the same. That’s how you avoid a depression – by stopping the death.
Mark Bahner
Apr 27 2020 at 3:43pm
It’s not “hysterical” to state that:
It is a simple fact. During the Great Depression, the peak unemployment rate was about 25 percent. The U.S. could have a peak unemployment rate higher than 25 percent if lockdowns continue for several more months:
“Soaring U.S. unemployment rates could approach Depression-era levels
The odds of any person in the United States who is under 50 years old dying from COVID-19, even if every lockdown was ended this minute, is probably less than 1–per-100,000.
I think you’re being very foolish if you believe what the Chinese government has told you about COVID-19 cases and deaths before and after the lockdowns.
CIA hunts for authentic COVID-19 numbers in China
robc
Apr 26 2020 at 12:12pm
That is exactly what we have been doing.
Todd Kreider
Apr 26 2020 at 5:08pm
“How is it even remotely possible to potentially reach “even millions” of COVID-19 deaths in the U.S., even if all restrictions in all states and localities are completely lifted tomorrow?”
It’s not and that was what was frustrating about listening to Justin Wolfers. I thought David Henderson made excellent points but at the beginning he said that the dramatic reduction in covid-19 deaths from 2 million to the currently estimated 60,000 has been in part due to the lockdowns “but some, and possibly a large part of it, is due to our private voluntary activity.”
That may be true, but the main reason we won’t have anywhere near 2 million deaths is because the quickly written-up internal paper that Imperial College released was likely wildly off to begin with as a worst case scenario.
David Henderson
Apr 27 2020 at 4:03pm
Actually I didn’t say what you attribute to me. I never bought the “2 million deaths without adjustment” estimate.
Todd Kreider
Apr 28 2020 at 11:10am
This is your quote from 8:05:
“Interestingly, Anthony Fauchi, his latest estimate a week or two ago, said he expects 60,000 Covid-19 deaths by the end of August, putting it in the same ballpark as our worst flu season. Of course some of that reduction [from Fuachi’s early estimate of up to 2.2 million American deaths] is due to the lockdown but some, and possibly a large part of it is due to our private voluntary activity. And that brings me to a key distinction that we have to make: social distancing and lock downs. Sometimes people equate them – they’re very different.”
You don’t mention that the millions of deaths estimate was almost surely way off. I just listened to your first rebuttal of Wolfer’s opening statement and you do say you never bought the 1 million deaths estimate.
Todd Kreider
Apr 28 2020 at 11:12am
That should be “Fauci” above.
Sean Kennedy
Apr 26 2020 at 8:16am
Excellent debate!
A few observations:
1) In the debate, DH questions the model of using the value of a statistical life (VSL) as giving strange outcomes as shown by LZ’s finding that it would be worth shutting down the country for 3 years. The reason this model doesn’t make sense is this context is because it’s not an apples to apples comparison. JW makes the claim that the value of the saved lives is 1M lives * $9M per life and compares that to the lost economic output. So he is comparing a subjective value of utility to an estimated real value of output. An apples to apples comparison would be to compare 1) the lost output of lockdown vs no lockdown and 2) the value of lives saved of the lockdown vs the impairment of the value of life due to the lockdown. Here we are now comparing apples to apples and the math should make more sense.
Back of the envelope calculation of the impairment of life due to the lockdown could be something like:
US Population 331M * VSL $9M / Life expectancy 79 / Days per Year 365 * Impairment factor (let’s say 50%) = Daily value lost of lockdown $52B
To make this even better it would be adjusted for the value of lost year in different age cohorts. While not a socially desirable thing to say I think most of us would agree that they would trade more than 1 year in their 90s for 1 year in their 20s.
As this is a daily value it should also be compared to the daily decrease in lives lost due to extending the lockdown as extending the lockdown has diminishing marginal savings of life.
2) On the cost of the lockdown side of the ledger JW only accounts for economic activity lost during the lockdown. What about the permanent decrease in economic activity going forward from an extended lockdown? What about permanent loss of utility for people who are unemployed, have permanent reductions in income, get divorced, etc…
3) JW made a claim about needing the lockdown because of externalities and how he wouldn’t be able to go to the park because he wouldn’t trust other people to keep him safe. This argument made me laugh because just because there is no lockdown doesn’t mean he has to go to the park! If he doesn’t trust people to take his safety into account he can stay at home where he would be if the lockdown was extended and his life isn’t changed except for being envious about other people having fun at the park! Why does he have to put his own risk assessment on other people? Also if you’re in a park and someone else isn’t being safe would have a negligible (arguable no) impact on your risk of getting the virus. Social distancing is a two way street! If someone is getting too close to you then you can move further away from them to maintain 6 feet. If you have an n95 mask you are probably going to be safe even if they do not. Vox has a good article on the risks of getting sick in the outdoors. A couple pertinent highlights are 1) less than 0.5% of outbreaks happened at an outdoor event and 2) to get the number of virus particles detected from a cough 8 inches away isn’t very high so the odds of having the required amount of virus particles 6 feet away with a mask, outside is miniscule.
https://www.vox.com/future-perfect/2020/4/24/21233226/coronavirus-runners-cyclists-airborne-infectious-dose
John Alcorn
Apr 26 2020 at 2:20pm
An observation by Robin Hanson seems cogent:
Michael
Apr 26 2020 at 2:57pm
I have the same feeling.
John Alcorn
Apr 27 2020 at 7:01am
Although we should take models with a grain of salt, see the new study by Dirk Krueger (UPenn), Harald Uhlig (U. of Chicago), and Taojun Xie (Nat’l U. of Singapore), “Macroeconomic Dynamics and Reallocation in An Epidemic” (NBER Working Paper No.27047). Here is an excerpt:
John Alcorn
Apr 27 2020 at 8:44am
Re: Justin Wolfers’ precautionary-principle concern about a risk of one million Covid-19 deaths in the USA.
See a new study by Joshua R. Goldstein and Ronald D. Lee (Demography & Economics, UC-Berkeley), “Demographic Perspectives on Mortality of Covid-19 and Other Epidemics,” NBER Working Paper No. 27043. (The paper is admirably clear, also for a lay audience.)
Here are excerpts, wherein the authors outline their approach and general findings, explore a hypothetical example of one million Covid-19 deaths in the USA, and contrive a useful hypothetical metric, “temporary aging”:
John Alcorn
Apr 27 2020 at 9:40am
A new study by Adriano A. Rampini (Duke U.), “Sequential Lifting of COVID-19 Interventions with Population Heterogeneity,” NBER Working Paper No. 27063, makes a case for a two-pronged approach. He would continue lockdown for the most vulnerable segments of the populace (by age, co-morbidity, and other special health risks), but allow the rest to return to work. Here are excerpts:
Again, we must take models with a large grain of salt. For example, Dr. Rampini’s model, for simplicity, assumes that workers are interchangeable.
Nonetheless, my intuition is that the Swedish approach is sound, if more effective measures are introduced to shield the vulnerable. Wouldn’t it be more cost-effective to focus attention and public resources sharply on shielding nursing homes (perhaps with strong incentives for workers to isolate there), and to allocate subsidies for the subset of working-age persons who have special health risks to shelter-in-place (with delivery of provisions etc)?
I am mystified that Neil Ferguson (Imperial College, UK) and other ‘influencers’ are defeatist about shielding eldercare facilities and vulnerable demographics via keyhole solutions (targeted measures). Ferguson argues that full lockdown is the only way to protect eldercare facilities until the general transmission rate is suppressed.
Jazi Zilber
Apr 29 2020 at 6:51pm
A few points stand out.
1. The argument that life years are irrelevant. Justin argues to value a deathbed life equally to a five-year-old. This is absurd.
It might be the case that some policy decisions ignore QALY. but this is an absurd yardstick. Just stickiness to some train regulations metric.
2. The $10m / life year is absurd in their own right, in my view.
Those are derived as I can gather from polling people with hypothetical scenarios. Daniel Kahneman famously promised to be able to frame those questions such that the values will be either X10 higher or 1/10 lower. Arrow told him that “we need a number, even if it is an absurd one” (free wording).
The NHS price to save one QALY is $35,000 while the UK safety rules use $70,000 per QALY. with the intuitive rational possibly to distinguish between those two criteria.
3. The average QALY lost per average COVID death is 5, according to some back of the envelope numbers here https://thecritic.co.uk/a-response-on-the-cost-of-the-coronavirus-shutdown/ (see the back and forth there if you wish).
4. The higher live prices were all decided from a luxury situation. Where it is basically how much society will spend from on some side issues to save lives. Nobody uses $10m per life (or per 90 year old life per Justin) on healthcare. Nor is anyone fantasizing on it even.
In fact, we are in this unique societal situation where suddenly the honest value of saving a life stands at the margin of the social decision. And some academic polling on how much would you spend to save a life becomes marginal to the real aggregate huge question that we have now.
5. Justin mentioned that some may remain with scared lungs for life. So the life/death binary is inadequate.
This is self evident in theory, of course.
But we do not have serious evidence that there a long term residue in large enough numbers. Speculative reports. But not much more.
Thanks for the debate and the blog
Jazi
Jazi Zilber
Apr 30 2020 at 9:30am
Another critical point is what is the target and end-game?
Taiwan, South Korea, China, Vietnam, and possibly others have aimed for a 100% no virus society.
This has long term value – assuming no entrance of foreigners infected, and perfect fact action of the occasional new cases.
The US does not even pretend to aim towards this.
This means that the implied lockdown is longer, and its expected value is much lower.
Blake Thompson
May 1 2020 at 10:52am
On the point about GDP, a simple way to calculate this is to look at Goldman’s annual forecast of -6.2%. 2019 GDP was $21.43tr. Thus, -$1.3tr for 2020 based on Goldman’s forecast. No need to get into the seasonality debate. Wolfer’s $500bn figure was cut in half, as you noted.
https://www.cnbc.com/2020/03/31/coronavirus-update-goldman-sees-15percent-jobless-rate-followed-by-record-rebound.html
Comments are closed.