The optimal duration of lockdown

Here is a new AEI paper by Anna Scherbina.  I have not read it and am not endorsing (or criticizing) its conclusions, here goes:

We investigate the optimal duration of the COVID-19 suppression policy. We find that absent extensive suppression measures, the economic cost of the virus will total over $9 trillion, which represents 43% of annual GDP. The optimal duration of the suppression policy crucially depends on the policy’s effectiveness in reducing the rate of the virus transmission. We use three different assumptions for the suppression policy effectiveness, measured by the R0 that it can achieve (R0 indicates the number of people an infected person infects on average at the start of the outbreak). Using the assumption that the suppression policy can achieve R0 = 1, we assess that it should be kept in place between 30 and 34 weeks. If suppression can achieve a lower R0 = 0.7, the policy should be in place between 11 and 12 weeks. Finally, for the most optimistic assumption that the suppression policy can achieve an even lower R0 of 0.5, we estimate that it should last between seven and eight weeks. We further show that stopping the suppression policy before six weeks does not produce any meaningful improvements in the pandemic outcome.

I hope there is a variable in the analysis for “risk of extreme societal pressures,” though I am not sure if those are higher for extreme lockdown scenarios or for extreme “let it rip” scenarios.  In any case, such risks are a significant factor in how I view these questions.

And now I must teach! (on-line, of course).  But I wanted to get this up for your attention before the evening closes.

Comments

Louis Buñuel, anyone?

Luis

But Tyler says humanities has taken a beating in the present humanitarian crisis.

And might even lead to a rise in status.

In the outpost of empire ?

The Monkey King lives on an island.

孫悟空?

At the risk of sounding mercenary, how does a virus that primarily afflicts elderly smokers, and the elderly with multiple comorbidities, pose a threat to GDP, except by the reactions to that virus?

a) It doesn't *exclusively* affect elderly smokers/people with comorbidities.
b) The resources required to care for those afflicted with the virus will exhaust the system's capacity. This could be avoided by taking a position like "anyone over 65 will not be treated for COVID-19" but that's unfeasible politically or ethically.

Why did we need a relief bill then (or did we)? The suppression policies saved us 9 trillion dollars! It seems reasonable to assume that at least some of those savings will be spent in restaurants in the future, so those restauranteurs should be thanking the government for increasing their future revenues. Instead of providing relief, shouldn't we be taxing those who have been "bailed out" of $9T of losses?

Another way to view it is that $9T for a population of 330M is $27k per person. So, if we had a vaccine right now and the company providing it wanted to charge $27k per dose, then no one would be right in complaining that such a price was immorally high because, after all, we would be saving just as much while avoiding suppression measures. It seems perfectly reasonable to charge an average family of four $100k+ for vaccination given all of the savings!

The thing is, that $9 trillion is the amount total US living costs aare expected to decline.

Isn't cutting living costs by $27K per person a great virtue to conservatives and the GOP who always talk about cutting costs?

After all, Trump is polluting the air more than Obama CAFE rules to cut US living costs, right?

He's advocated burning more coal to cut living costs.

Opposed Obsmacare to cut living costs.

Economies are zero sum. Your cost is my income, and my cost is your income.

Businesses can't get more income from consumers than businesses and government puts in consumer pockets.

And AOC and Trump both call for the Fed lending money to People who will never repay it (government is We the People, corporations, like government, is People.)

Yeah, I know, Soylent Green, spoiler alert, is People.

The fact remains, the vast majority people requiring ICU are the elderly. The workforce was and is hardly affected at all. The dent to GDP comes draconian government ukases, not from the virus. (Oddly, yesterday's "libertarians" are today's statist martinets. In every libertarian's closet is a uniform shirt, jodhpurs, high boots and epaulettes.)

We could build tent hospitals and manufacture ventilators in a hurry, for $X billion, and sequester the elderly.

Or we could sequester everybody, shut down the whole economy, and still build tent hospitals and hustle up ventilators.

I hope I am wrong, but the consequences of the lockdowns could be catastrophic. Widespread rent strikes. Food lines. A collapse of the financial system. More suicides, drug use.

Once renters in NYC and the West Coast get used to not paying the rent...how many renters are there vs. county sheriffs?

And how does one end lockdowns as long as the population is still naive to the virus?

Has anyone an exit strategy?

Digging out of the Great Recession took 10 years. Greece never did dig out.

And they remain that the majority of ICU cases (which is not equal to the number put on ventilators) are not the elderly. In most countries, the majority of deaths involves the elderly, who generally die much more quickly than the young put on ventilators.

And Germany, with fairly good testing and quarantining, contradicts this attitude, "The workforce was and is hardly affected at all." Covid19 is not an old person's disease (yet), which is the major explanation for the low death rate.

Deaths/1M pop - US 12, Germany 9 (source=worldometer) Just a couple of days ago, the U.S. rate was 4.

"I hope I am wrong, but the consequences of the lockdowns could be catastrophic. "

The consequences of unchecked pestilence could also be catastrophic.

"Once renters in NYC and the West Coast get used to not paying the rent...how many renters are there vs. county sheriffs?"

Dead or sick renters don't pay rent. Dead or sick sheriffs don't enforce the law.

Have a little faith, BC. These lockdowns aren't theory. It worked for Asia. Take their best ideas and run with it. If you think world leaders made a serious mistake then use the opportunity to devise workarounds for the next serious global crisis because this one is the template for the future. Save the chicken little stuff for when you get really bored and not when you are particularly anxious.

You are incorrect to claim that the US version of lockdown is what "worked for Asia". What worked was a combination of >forcedpossiblynone< of that.

1. Forced lockdown. 2. Universal (regional) testing of those working of traveling outside the home. 3. Aggressive contact tracing of positives along with quarantines of those who *might* have been exposed. The US is doing none of that.

More likely we don’t have the state or civil societal capacity to do real lockdowns....so we will have ineffective half measures that as a byproduct cause 33% unemployment.

So we’ll get the worst of both worlds. Mass infection with a side of Great Depression 2: Corona Bugaloo

"We furthermore assume that only 90% of the infected people develop symptoms that would need to be treated with medications, missed work, medical visits, or hospitalizations."

Just pointing that out.

Everyone (and their dog) seem to have a covid-19 model these days, so here is mine for the UK

https://www.kaggle.com/kiwiakos/covid19-uk-model/

Their estimated incremental economic cost of suppression (relative to mitigation) is about $36 billion a week. Their scenario where suppression reduces R0 to 1 is lifted after 34 weeks and is then substituted with a mitigation policy for another 34 weeks. I couldn't find where they discussed the costs of mitigation as well. It seems like their economic cost estimates are too low. And I think Table 2 (giving the age profile of the US population and the age profile of morbidity of Covid) shows why VSL is the wrong metric and the value of life years saved should be used.

This looks more and more like the most disastrous response to a disease in the history of the planet.

We’ve blown up the entire economy of the planet for what looks to be a somewhat bad flu virus in impact.

Yeah, kind of. But as a flu, it really isn't that bad.

At least considering how covid19 is treating them - This looks more and more like the most disastrous response to a disease in the history of the planet.

"We’ve blown up the entire economy of the planet "

No we didn't. I went to the store today to buy milk and eggs. I just proved you wrong.

Flu viruses don't kill 2.2 million people in a few months in the absence of efforts to suppress it. At least no flu viruses in the last 100 years have done so.

Good luck running an economy with 2.2 million dead bodies piling up. Thanks to the suppression methods, we *might* get lucky and get 100 to 200,000 deaths in 2-3 months, still VASTLY more deadly than the worst of of the recent flu seasons (which run 9 months and do not overtax our medical system).

Models don’t kill 2.2 million people either.

Truth is, we probably already have >50% exposure in the hard hit areas. Deaths will be trending down.

We’ve destroyed the global economy for a virus likely to kill at most 1-2 million people globally. The economic depression will almost certainly kill far more.

At least regarding "Truth is, we probably already have >50% exposure in the hard hit areas. Deaths will be trending down." And the people hoping that are living in Italy and Spain, where the idea of things trending down has become something of a tired joke in Lombardy or Madrid. At some point, yes, things will be trending down - at best, the rate of increase is trending down. Which would be nice, bless the virus's little heart.

Sadly, the U.S. has a long road of trending up ahead before it starts to actually go down the other side. Assuming you are not a victim of TDS, as Trump is currently saying a range of 100,000 to 240,000 dead is realistic, while adding that such a low toll would be proof of how well his administration responded.

"Trump is currently saying...". Ah, you're one of those good people who rely on our President for their facts. Sad or laughable, take your pick.

https://www.rep-am.com/life-arts/family/2020/03/21/coronavirus-cold-or-the-flu-heres-how-to-tell-the-difference/

Coronavirus symptoms typically start gradually include the following, according to WHO:

Symptoms are typically mild, according to WHO, with the CDC saying that is also typically the case with children. Sometimes, people become infected but do not show any symptoms or feel unwell. One in six people infected with COVID-19 falls seriously ill and experiences trouble breathing. However, older people and people with underlying illnesses – including heart issues, diabetes or high blood pressure – are more likely to develop serious illness.

----

Say what? One in six of the infected get seriously ill? How do they know that, my observation is that number is still bouncing around quite a bit. I am hearing some remarks from experts that rely on unstable data.

OK, NYC seems to fit. Monday they had 9,000 new cases and 1400 more hospitalized, that is close enough to one in six, given the stability. So on NYC five of six are in home isolation. Of the ones tested who would have mild symptoms,probably got a test voluntarily. So of the one in six who survive they go home to clusters where five in six are immune.

A winding tautology to translate assumptions into a different language.

Exactly. Why using mathematical language when you can explain your ideas in plain english ?

But what is the optimal lockdown duration when R_0 = 1.4?

https://www.nytimes.com/2020/03/29/us/seattle-washington-state-coronavirus-transmission-rate.html?action=click&module=Spotlight&pgtype=Homepage

Good Link. And here they are happy its down to 1.4
Seattle’s mayor, Jenny Durkan, ------------------- expected distancing requirements to continue in some form for months.

If you dislike TC and this site so much why do you spend so much time on it ? Masochistic, are we?

WA state is still growing quickly, though. Around 25-th of March, the state had it's best 24-hour period of new case growth--just 4.5% (16 day doubling). The next day is bumped back up to 24%, then 16%. Currently the latest data has it around 11 to 12%, which is 6-7 days for doubling.

It needs to be 12+ days to let the hospitals see numbers in beds reduce. So there's a way to go still.

The worst day for WA was 11-Mar, with 64.8%, and 14-Mar, with 40.5% increase.

Tech companies mandated work from home around March 6, and schools shut down in countywide a weeks later. So, we've seen the results of every effort taken thus far pay off. There arent' more events put into place that will delivering another big drop in the coming weeks.

I partially read the thing. It seems they want to spread out the infections until we get medicine but before restrictions cost too much:
"As discussed in Ferguson et al. (2020), the suppression policy helps reduce the total attack rate
(AR) (defined as the fraction of the population that becomes symptomatically ill) by shifting the
pandemic curve further out in time and closer to the vaccine availability date"
-----
Everything seems to depend on the arrival of a vaccine. The cost of shutdowns need be replaced with the cost of idiosyncratic suppression, differentiating between someone plowing corn and someone in church. Suppression costs vary all over the map. Simply banning mass gatherings for a while gets you half way there. Add in good masks and other social distancing depending upon the workplace.

With no interventionsd:

"Figure 1 presents the shape of the COVID-19 pandemic curve in the U.S. without any interventions. The number of new infections peaks around the middle of July 2020. The attack rate (AR) is 79.5%, and the total number of deaths is estimated to be 1.9 million."
..
"The economic cost of the pandemic, calculated as described above, is projected to be $13.2 trillion when VSL for the fatalities is taken into account and $1.2 trillion when it is not.
---
VSL is the value of a statistical life. 3.8 million of us die each year, at that rate we spend 26 trillion a year on dying, which is mostly what we do in economic I guess, prepare for death.

But the shutdowns reduce deaths over all, and they need to include that with a VSL also, or leave VSL out of the picture.

Let us leave VSL out of the picture and we see direct costs, mostly medical, about 1,.2 trillion; about half the size if the ;stimulus'; which, by conclusion, we can call a bailout.

My answer? Keep scrambling about and trying different form of treatment and suppression. consider what you need, individually, for your work or household to keep R0 down. Get more stable data and re-run the simulation.

The hump to equilibrium in the paper has a sigma of about eight months. We are a third way through, on average, and New York about half way through.

Let us include Statistical Life costs of 13 T for 1.9 million, wrong but it cancels. The shutdown saves lives which otherwise would have died, as stats show. So the optimum, using that stat, says the duration where costs even will be mostly where the number of deaths saved from shutdown equal the number of deaths from deaths lost to those miss the vaccine. The stat is wrong, like I say. That stat says we are all mostly busy dying.

Is this a haiku?

The last line of the quoted extract is a flat out lie.

Sometimes, the trolls really are too lazy

Economics Spring 2020 ECON 846: Industrial Organization and Public Policy II Instructor Tyler Cowen He is also teaching ECON 104-DL1: Macroeconomic Principles this semester. Likely using one of the most self-recommended economics textbooks ever written.

At least get your post deleted for facts, and leave the misrepresentations to the professionals who care about nothing except influencing others.

On the other hand, if you mean TC is not actually teaching online, that is completely beyond my ability to easily check using google. Meaning maybe you aren't posting a flat out lie, just a half-truth or something with only an element a minimal level of truth.

Am I reading correctly that they assume the value of an 88-year-old life is (roughly) the same as the value of an 18-year-old life?

As a note of comparison to the model in this paper.
In Wuhan after lockdown ( Jan 23). It took :
12 days for the daily new cases to peak ( Feb 4)
39 days for the new cases to go under 100 a day ( Mar 02) = 5.6 weeks
76 days for the lockdown to end April 8 = ~10.9 weeks
By then Wuhan will have been more or less 3 weeks without a native case ( all imported).
Italy started national lockdown on March 9, new cases peaked on Mar 26 ( 17 days).
The US started lockdowns on Mar 19 ( CA) ,March 22 ( NY) and March 23 ( WA) for the the most affected states. New cases have not peaked yet
If we continue for 39 days ( May 1/ Trump's recommendation) and we're close to the Chinese scenario ( serious confinement, use masks) , then our curve would look like the Ro =0.5 case in this model ( Fig 5/week 7)
Infection returns in the fall but we have time to :
-mostly avoid hospitals overwhelmed
-develop better treatments
-perfect contact tracing.
I am not too optimistic about a vaccine this year.
The Chinese however quarantined the infected/asymptomatic in separate facilities ( not at home with family). This might have helped bring Ro down quicker.
In the real world the infected/ ever asymptomatic are probably 20 to 50% ( see Diamond Princess and Iceland DEcode) but are infectious for 3 to 5 days . This is why they need to quarantine separately.
If this model is correct ,we may have chosen already a (fairly) optimal strategy

But can we trust the data from China? I don’t usually where a Tin-foil Hat, but the “evidence” of China not being forthright is pretty damning.

Perhaps, Korea is the better litmus test?

And I wonder why Italy peaked so quickly from the date of full quarantine? Probably says more about the quarantines lack of effectiveness than the alternative.

MR knows excellence, this being an excellent demonstration of mood affiliation in action - "I have not read it"

Another “paper’’ that uses the full VSL for each death instead of the value per month of life remaining.

Likely this is done since applying the correct methodology reduces the cost of the disease by ~ 10x and this vastly changes the optimal policy.

And fails to look at the cost of lockdown in the form of "shoeleather": I lost a business and it takes me 6 months to repoen, recession related suicide and drug use, unemployment, etc. Fails to look at cost of a 2 trillion dollar bailout and giving more power to the government. Fails to look at cost of business that are open but have made huge investments in WFH software. And fails to look at the fact that a global pandemic was entirely predictable and hospitals should have been prepared with ventilators and masks and Tylenol and we should be suing hospitals and their administrators for negligence and fraud for the 2 trillion

To check their working, impute into the same model the savings from lockdown via the larger set of circulating infectious diseases which cause death at lower frequency than Covid19 but which share the same mortality profile (vast majority old), for which vaccination is not effective in preventing them and which lockdown cuts just as much.

Perhaps they'll find the vast majority of Americans are actually permanently worth more locked in and getting mailed a UBI cheque out of the "savings" that them not going anywhere leads to. Or maybe something will be wrong with their model....

And reading is fundamental, as Americans of a certain age recognize. The ones more likely to die, the ones who first associated RIF with a more positive meaning.

A few things learned while reading this one:

1/ the assumptions for R0 under suppression (0.5, 0.7, 1.0) all seem unrealistic for USA - hell, we're still haggling over whether surgical masks help attenuate virus spread let alone producing enough for our health workers

2/ the assumptions we're using are still Wild Ass Guesses (WAGs) but seem to be trending to Sophisticated WAGs - models are only as good as the inputs and we're failing hard here. My best guess is that this failure "bug" is deemed a "feature" by cynical politicians and do-gooders alike ("let's dupe the public into (choose one: longer suppression period, shorter suppression period)" Plug for joining the #surveillanceMedecine movement -> better model inputs -> better policy decisions -> less costly pandemic. Plug to share ZIP Code, e-mail address, and whether you have a fever each day with https://covid.joinzoe.com/ !

2b/ Recall Warren Buffets 2009 annual letter to BRK shareholders; quote "beware of geeks bearing formulas" -> what are the big variables being left out of the model?

3/ the model uses Value Statistical Life (VSL) measure to quantify costs of pandemic-caused death, which is something akin to "how much are you willing to pay to avoid death?" This seems wrong to me, as it should reflect years remaining I.e. younger deaths should be treated more costly all else equal. Seems like it was used to avoid the icky-ness of this debate, though using e.g. Quality Adjusted Life Year (QALY) -based measure may in fact change the optimal policy response A LOT

4/ older, sicker deaths may be less costly than the middle-aged healthy. 80 y.o. with Diabetes are more likely to die quickly or not be treated at all, whereas a % of 55 y.o.'s will need ventilator + a week in ICU. Whether this more supports or undermines those who favor a "no suppression, let it rip" response was not clear to me. Hot take is that adjust model for QALY would further undermined the "let it rippers" but it's a pure guesswork

The let it rip folks are social distancing, they are getting medical treatment when they are sick, etc.

Stated vs revealed preferences.

They want everyone else to let it rip to avoid financial loss while they bunker down to save their life. It’s almost criminal.

Thank you. Like everyone nowadays that uses 'virtue signalling' as a criticism--unreliable. Bit more at stake here, though.

How do they model Japan, doing fine with no lockdowns?

Watch Japan over the next two weeks. It's only fine if you are a dog in a fire.

Comments for this post are closed