Basically the Atlas et. al (Hill) piece is a flawed comparison because it omits the number of life-years lost in the counterfactual scenario of no lockdown.

Here's what a World Bank study has to say, after quantifying the poverty results of the lockdowns, which are actually higher (in relative terms) in developing countries compared to the developed countries:

For most developing countries, one would have to hold unrealistically high values of α (“I’d rather spend 600 years in absolute poverty if that would add one year to my current life expectancy”) to judge that the mortality effects dominate the poverty effects as sources of welfare losses. (Despite the fact that our approach treats the value of human life exactly the same way across all countries.)

That is not to say that a laissez-faire, “no-intervention” response to the pandemic would have been superior [... For that] one would have to simulate a counterfactual scenario where governments had not responded, and the epidemic evolved until countries (hypothetically) reached herd immunity.

We constructed such a counterfactual scenario using the Banerjee et al. (2020) assumption that herd immunity is reached at an 80% infection rate and found that such a no-intervention policy would have led to higher welfare losses, particularly in richer countries. For high income-countries we estimate that the number of lost years under this scenario would have been five times larger than the sum of lost years and poverty years estimated as of early June (based on the $21.70 poverty line). This implies that the welfare consequences of no-intervention in these countries would have been at least 5 times worse than the consequences measured as of early June. (Five times is based on α =1. The ratio obviously rises with larger values of α.)

---

As for the Hill piece, it is authored by a number of academics... but, the lead author is affiliated with the Hoover Institution, and more importantly they haven't actually published their "model" anywhere but in opinion pieces, as far as I can tell. As whom cited (or republished) them, the list speaks for itself: AIER and Mises Institute have written echoing pieces based on that. (And it's not the case that there haven't been plenty of economic models of the lockdown-related trade-offs published in economics journals.)

As for other factoids from the lead author of the Hill piece: "99 percent of [Covid-19] infected people have no significant illness from it." I'm curious to know for what definition of "significant" that statement is true.

The sums may be correct; but the number calculated is utterly meaningless.

The primary impact on deaths of the shutdown is the number of years of life lost is those saved by the reduction in number of Covid-19 cases and thus deaths. In reality this number almost certainly dwarfs every other number used in the calculation since without the shutdown cases, and thus deaths, would have carried on growing exponentially until a sufficient proportion of the US population had caught the disease for herd immunity to kick in leading to an estimated number of deaths of well over a million (given an IFR of 0.7% as seen in France, which would likely be an underestimate in these circumstance). This number of deaths, by the author's method and assuming the same age profile, would lead to over 10 million lost years of life

What they have done is like trying to assess the effectiveness of car safety belts and counting only the lives lost by people despite wearing safety belts and not considering the number of lives saved by wearing seatbelts. It's a total nonsense. By leaving the most important number out of the calculation, the source has calculated a number that is completely meaningless and whether or not the actual numbers add up is essentially irrelevant.

Because of this hopelessly inadequate calculation the conclusion derived is completely unsupported and we cannot conclude that the shutdown will cost years of life. In fact, given the scale of the missing number, it's probable that it has saved millions of years of life.

Maybe, but this article doesn't show that. It is deeply flawed. It is a reiteration of the "sacrifice the elderly" argument with more statistics.

The article uses the incorrect points of comparison in various ways, it presents a rosy picture of their proposal that was known to and is being demonstrated to be false, it fails to demonstrate their plan is better than any of a number of alternatives, and as a public policy proposal it fails to examine its moral and ethical ramifications.

The US Did Not Lockdown At All Costs

The article opens with their view of the situation in the US.

Our governmental COVID-19 mitigation policy of broad societal lockdown focuses on containing the spread of the disease at all costs, instead of “flattening the curve” and preventing hospital overcrowding. Although well-intentioned, the lockdown was imposed without consideration of its consequences beyond those directly from the pandemic.

The piecemeal US federal, state, and local government response certainly did not try to contain the spread of disease "at all costs". Particularly the lackluster, contradictory, and often harmful federal response. As well as some state governments which were in denial and are now seeing spikes.

Governments who did take COVID seriously did focus on flattening the curve and preventing or reducing hospital overcrowding. And they did take into account its consequences as evidenced by economic boosts to unemployment insurance, suspending evictions, "small" business loans and grants, and so on.

Because of the piecemeal response the US did neither "lockdown at all costs" nor "flattening the curve". What we're seeing is the result of a mish-mash of everything from full denial to full lockdown with changing and confusing recommendations and mandates. While other countries similar to the US have COVID in steady decline, the US is on a sharp rise again. If we look at its success in similar countries, the "broad societal lockdown [focused] on containing the spread of the disease at all costs" they claim they're comparing against looks very different from what the US has done.

They're not comparing their reopening plan against a "broad societal lockdown", they're comparing against a broad societal screw up.

Their Plan Is Already Backfiring, and the US Knew It Would

To end the loss of life from the economic lockdown, businesses as well as K-12 schools, public transportation, parks and beaches should smartly reopen with enhanced hygiene and science-based protection warnings for any in the high-risk population. For most of the country, that reopening should occur now, without any unnecessary fear-based restrictions, many of which repeat the error of disregarding the evidence. By following a thoughtful analysis that finally recognizes all available actions and their consequences, we can save millions of years of American life.

The article was written May 25th, 2020 when US COVID cases were at their lowest since its peak in early April. Many places did as they suggested. Now our case load is spiking again.

Source: Washington Post

Clearly it is not working. And it's what Homeland Security predicted in March 31st.

Steady State assumes school closures until summer, 25% of people telework, and there is some social distancing

The Steady State + 30 Day shelter-inplace scenario has a greater rebound peak after the mitigation is relaxed (assuming further shelter-in-place policies are not implemented to reduce futurepeaks) than steady state

We can also look at Sweden. Sweden never locked down, how are they doing? Their death toll is at 500 per million, one of the highest in the world. Their case load continues to rise.

Comparing Apples To Apple Trees

The past century has witnessed three pandemics with at least 100,000 U.S. fatalities: The "Spanish Flu," 1918-1919, with between 20 million and 50 million fatalities worldwide, including 675,000 in the U.S.; the "Asian Flu," 1957-1958, with about 1.1 million deaths worldwide, 116,000 of those in the U.S.; and the "Hong Kong Flu," 1968-1972, with about 1 million people worldwide, including 100,000 in the U.S. So far, the current pandemic has produced almost 100,000 U.S. deaths, but the reaction of a near-complete economic shutdown is unprecedented.

Here they're comparing the complete toll of historical pandemics with the toll so far in the first few months with the existing measures and advances in medicine. The COVID death toll is likely understated. It is not an apples-to-apples comparison.

For example, the Spanish Flu lasted two years and had three major peaks. One early. A much greater one after four months. And a middle one after eight months.

The US is only about four months into the COVID pandemic. They've had their first spike, and before the first wave even finished the second one is starting. What did the Spanish Flu look like four months in?

Source: Wikipedia

And let's look at the US case load four months in again.

However, deaths have not spiked (the June 25th spike is a statistical glitch). But deaths lag cases; it may still be too early to see their effect.

After four months we're seeing a second spike. If the US follows the pattern of the Spanish Flu, and I hope they do not, this is only the beginning. Thus one cannot simply project the current state of affairs on May 25th forward when comparing with historical pandemics.

The Counter-Factual Fallacy

Jack Aidely summed this up very well.

What they have done is like trying to assess the effectiveness of car safety belts and counting only the lives lost by people despite wearing safety belts and not considering the number of lives saved by wearing seatbelts. It's a total nonsense.

The danger of "what if" scenarios is two fold.

1. Making your change while assuming everything else remains the same.
2. Giving your scenario every break.

To the second, as we are seeing, what looked like a rosy reopening scenario in late-May has turned into messy spike in cases.

The authors compare the actual (and likely underestimated) effects of COVID with their estimated effects of a lockdown. The authors are trained in their fields, but biases towards their preferred scenario will appear.

To the first, they're comparing the (probably underestimated) deaths under piecemeal lockdown to the potential deaths due to economic inactivity. But if the US did not lock down, their cases and deaths would have been different; given the current spike they likely would have been even higher.

To have anything like an honest answer, they must compare the current scenario with a no-lockdown scenario. Homeland Security predicted at the end of March 300,000 deaths in an "umitigated" scenario, and 200,000 deaths if we reopen too early. I wonder what their models tell them now.

The article cites things like "lives also are lost due to delayed or foregone health care imposed by the shutdown and the fear it creates among patients", but an even greater pandemic would have produced an even greater delays and fear.

There is the assumption that if there was no lockdown, everyone would have proceeded as normal. Again, we know that not to be true. Already people are reluctant to risk their lives, or the lives of those around them, to go back to work, or to get a haircut, or go to a restaurant. In an even worse pandemic caused by staying open, the economy would still be impacted.

Finally, there is the emotional toll. The article's "years of life" is really cover for "sacrifice the elderly" and the immunocompromised. I have family and friends who are immunocompromised. Were I to get infected, they might get infected and could die. The thought that my family or friends died because I wanted a hamburger is horrifying.

The Demographics of COVID Patients Is Changing

The article assumes COVID will continue to target mostly the elderly once the economy reopens. But this may not hold true, and there are already signs it is shifting as states reopen.

Houston Methodist Hospital in Texas is seeing a shift towards younger patients. Their president and CEO Dr. Marc Boom says

"What it tells me is that the older, more vulnerable population is staying home and staying safe and the people younger who said, 'Hey, you know what, this is behind us, this is a lot of hype,' they let their guard down, they went out doing life as usual, and we're paying the price for that," Boom said.

Texas Gov. Greg Abbott commented when putting lockdown restrictions back into place that they're seeing more young people testing positive.

"The majority of people who tested positive since the beginning of June have been people under the age of 30," Abbott said at a press conference last week, adding that some of the increase was a result of "bar-type settings."

Source: NPR

If as the US reopens COVID begins targeting younger people, their whole argument collapses.

The Chosen Metric Devalues the Elderly

The article cannot be taken as a naive actuarial exercise, nor can we answer the simple text of the question and then stop. This is not an actuarial paper publishing their results in an academic journal. It is a public policy recommendation by public policy professionals with real consequences for real people. We must examine their chosen metric and its consequences.

In making their "the cure is worse than the disease" argument, they choose not "deaths" but "years of life" as their metric. Since a high proportion of COVID deaths are among elderly people, this allows them to paint a rosier picture of their deaths. They have devalued the lives of the elderly; you're worth less if you're older. And they've done it on a simple linear scale.

While it is valid to take such actuarial values into consideration, they should be one of many in a policy argument. Instead, the article focuses solely on years of life. Reducing one's value down to a single data point is fraught with ethical problems which the authors do not examine. Nor do they disclaim the article as an actuarial exercise not to be taken as a policy position. A position which rests upon the idea that one person's life is more valuable than another's must be thoroughly examined.

The Trolley Problem

The authors are proposing a solution to The Trolley Problem.

The Trolley Problem is an ethical thought experiment. The classic formulation is you're at the controls of a trolley speeding towards a crowd. You can change tracks to kill one person instead. What do you do?

It has infinite variations, as The Good Place illustrated in grisly detail. The answer seems obvious, you minimize casualties. What if we reformulate it? Now you're a doctor choosing between saving their patient, or letting them die and using their organs to save many others. Now what do you do?

Medical professionals have their own very real Trolley Problem when they are required to perform triage. They do this only when the need is acute and immediate. Their decision making process is backed up by a large body of ethical considerations and guidelines. To avoid conflicts of interest and biases, their metrics are about the immediate medical situation: the severity of the condition and the likelihood of recovery.

The ethical danger of the author's solution is they are triaging based on their perceived value to society, and they're doing it to an entire population. They are the doctor who will let their patient die without their consent so they can harvest their organs to save someone of a group they consider more worthy.

You Can Sacrifice Any Group With The Right Metric

By choosing the right metric, one can justify sacrificing any group they wish. If one chooses a person's contribution to the GDP, they could justify sacrificing the unemployed and disabled. If one values certain industries more than others, they can justify sacrificing the workers of the "lesser" occupations. Senator Tom Cotton recently argued that someone in Wyoming who does "mining, logging and construction, and... manufacturing" has more right to representation than someone in DC who are "bureaucrats and other white collar professionals".

False Dichotomy

Which brings us to the real problem with the "lockdown is more harmful than COVID" issue. The article presents it as if we have only two choices: your money or someone else's life.

The article cites things like "every $10 million to $24 million lost in U.S. incomes results in one additional death" and "lives also are lost due to delayed or foregone health care imposed by the shutdown and the fear it creates among patients". I don't doubt these claims. What I do doubt is that there was no way we could have mitigated them.

Why is the US doing so much worse than comparable countries? How did those countries do so well? What could the US have done better? What can they still do?

Why is the health and welfare of the US population so tied to their jobs? How does that compare to similar countries? What could the US have done to mitigate that before a pandemic hit? When the risks became clear back in February? What can they do now that we're in the middle of it?

Why was there so much fear to seek health care? The chaotic and contradictory response probably had a lot to do with that. The nature of US health care and health insurance is another. What could we have done to ensure that people did seek health care?

The public health solution to the COVID trolley problem is to slow the speeding trolley down: don't overload the medical system in the first place. These are measures to flatten the curve: husband medical resources where possible without impacting critical care; expand access to medical care; expand production of vital medical resources; and get as many healthy people to wear masks and take as many protective measures as possible.

That last one includes enacting public policy which ensures people do not have to choose between staying safe and paying the bills. Examples include: expanded unemployment insurance, moratoriums on evictions, expanded ability to conduct business at a safe distance, expanded health insurance not tied to your job, and so on.

The article fails to consider this as a third option, or any other options. Considering its grisly calculus, it is imperative to do so.

"Reopen and sacrifice the elderly" is not the only solution.

Article claim: Specious on its own terms

Contra the excellently researched answer from Paul Draper, the claim in the article is untrue at the time it was written.

The article's central claim is that the COVID-19 lockdown has led to 700,000 years of lost life per month, of which 200,000 are due to rising mortality from unemployment and 500,000 are due to missed preventive/maintenance healthcare. They count this against 800,000 lost years of life directly from COVID; let's call that roughly 200,000 lost years per month during the quarantines (March to late-May).

The article authors' argument is incorrect because they misattribute the 500,000-lost-years-per-month from foregone medical care to lockdown, when it is actually due to the disease itself.

Health care access has never been prevented by quarantine measures. There have been no bans on traveling to health care appointments. No hospitals or doctors' offices have been closed as "non-essential services." To the extent that non-COVID-related health care is being foregone, it isn't because of "quarantine"--it is because the disease itself saturated health-care systems with COVID-19 patients and because accessing health care during a pandemic comes with a vastly greater risk of exposure to a highly infectious deadly disease.

Those 500,000 years lost per month due to missed healthcare do not belong in the "lockdown" column; they belong in the "COVID-19" column. The authors should be comparing their 200,000 years-per-month from unemployment to 650,000 years-per-month from COVID-19 and its impact on the capacity and perception of the health system. If you fix this accounting error, the lost life due to COVID clearly exceeds the lost life due to the quarantine itself by a factor of three or more.

Note that through all of this I am merely stipulating that the authors' numbers are correct. They have not considered excess vitality due to lockdown, such as the 88% drop in fatal car crashes during California's lockdown, the impact of reduced pollution, etc. I don't blame them--these are difficult to measure--but not more difficult than the predicted excess-mortality effects of unemployment. This accounting enterprise requires much more careful research than an op-ed.

Future

Of course, the article was written in late May 2020. As of late June 2020, having ended the lockdowns, the US is adding a whole lot more points to the COVID-19 column, with back-to-back days setting records for new infections. Any accounting-based approach to the cost of quarantine and of COVID will have to wait until the pandemic has actually ended.

Edit: Sources and Extended Discussion

I'm adding the following section in response to requests for sources, and to make sure that the above is read clearly. I will try to respond to any further requests for clarification rapidly.

Background:

Source article claims: We are responding to the article in OP, an opinion piece published on 2020-05-25, which I will refer to as Atlas. The central argument of this piece was that government (federal/state/local not specified) 'shutdown measures' (precise details not specified; I infer this means business closures and stay-at-home orders) up to the publication of the article were overly restrictive. To support this argument, Atlas et al claim that the 'shutdown' would be responsible for more lost life-years per month (on an ongoing basis) than the disease had been. This specific claim is the subject of the question.

Atlas et al calculate 700,000 life-years lost per month due to shutdown, of which 200k are due to 'deaths of despair' from unemployment/economic hardship, and 500k are from deaths due to postponed health care. While Atlas et al make no effort to separate economic damage due to quarantine measures from economic damage due to the coronavirus itself, for the purposes of this response I stipulate that number to be correct. I argue that their claim is untrue because the 500k life-years from postponed treatment are due to the coronavirus and not to quarantine measures.

Definition of "government response": Atlas makes references to 'governmental COVID-19 mitigation policy of broad societal lockdown', 'economic lockdown', etc. but does not define them. For the purposes of this answer, I am taking these terms to mean 'enforced action by government agents to require the closure of a business or deny provision of services that would have been offered absent the government action; or to limit the movement of individuals.'

One of the vexing properties of commenting on US pandemic response is that it has been largely delegated to state and local actors. There was no federal lockdown, and probably couldn't have been anyway. A complete survey of state and local lockdown measures is not practical in this forum. I will consider the response in New York (city and state) as representative, both because it's what I'm most familiar with, and because (as the hardest-hit region at the time Atlas was written) its lockdown was likely among the most stringent.

Timeline of NYC and NYS response, and selected federal guidance:

These are also summarized by Citylab.

• City Quarantines:
  • Executive Order 100 (NYC100) -- March 16th, 2020. First imposed NYC quarantine measures were imposed for a period of five days starting on March 16, 2020
  • Executive Order 102 (NYC102) -- March 20th, 2020. Extended NYC100 for a further five days and issued some clarifications. A further series of executive orders has extended shutdown measures in roughly five-day increments since.
• At the state level:
  • NYS202.3, March 16th, 2020. Bans public gatherings (concerts etc) over 50 people and prohibits in-person dining/bars.
  • March 17th, other measures were taken to allow out-of-state medical providers to practice; to shut down schools and adult day-care centers, shopping malls, etc.
  • NYS202.6 (March 18th, 2020), which mandated non-essential businesses to reduce staff counts to 50%.
  • NYS202.6 was followed by the main "lockdown" order, the NYS PAUSE Plan (PAUSE), signed into effect March 20th, 2020.
• CMS Recommendations on Non-Emergent Elective Services (CMS) of April 7th, 2020
• CDC Recommendations to Healthcare Providers as of 2020-05-12 (CDC-Prov)
• CDC advice to individuals (CDC-Indiv) shows the current version. Wayback Machine shows that at the time Atlas was published, the recommendations were substantially the same and had been since May 12th, though a section about recommended vaccinations was removed in the current version. (I've linked the current version first because the expansion-card feature is broken in the Wayback version; you have to view page source to read it.)

Argument

Atlas claims that "lives also are lost due to delayed or foregone health care imposed by the shutdown and the fear it creates among patients." This one sentence is Atlas' entire justification for including 71% of the life-years which they blame on the quarantine. Examples cited in Atlas include evaluations for strokes, missed chemotherapy treatments, non-diagnosis of new cancer cases, reduced living-donor transplants, and reduced childhood vaccinations.

Based on the above and other sources, I argue that while patients are limiting access to health care, it is not due to quarantine measures. Specifically:

• Quarantine measures did not include shutting down health services, and in fact individuals were encouraged to continue receiving health care. To the extent that health services were not available, it was due to the COVID-19 disease itself and to choices by non-government actors to reduce risk of exposure and spread of the disease.
• To the extent that individuals delayed care out of fear, there is no evidence that this fear was prompted by government mandates. The Occam's-razor explanation is that people were afraid of COVID because COVID is dangerous, not because the government implied they should be afraid.

Since this excess mortality cannot justifiably be blamed on the quarantine measures, the only remaining option is to blame it on coronavirus itself.

Point 1: Did government quarantine measures prohibit access to the kinds of health care cited in Atlas?

No. NYC102 specifically clarified that "Any essential business or entity providing essential services or functions shall not be subject to the in-person restrictions [prohibiting in-office work]. This includes essential health care operations including research and laboratory services". The same language appears in NYS202.6. In fact, guidance on NYS202.6 as early as March 25th, 2020 clarified that this included "walk-in-care health clinics and facilities; emergency veterinary, livestock services; doctor and emergency dental; licensed mental health providers..." by April 9th, it was further clarified to include emergency chiropractic services, physical, and occupational therapy. Moreover, the PAUSE order's 10 points included the statement that "Sick individuals should not leave their home unless to receive medical care [my emphasis]" -- far from denying care to patients, people who believed they were sick were specifically told they were permitted to seek care.

What the NYS orders did do was cancel or postpone elective surgeries. However, hospitals had already been doing this voluntarily, as reported in the NYT, byline March 14th (two days before NYC100 made it an official directive). Of course, some hospitals kept doing them anyway, without penalty--because there was no federal mandate on this, only voluntary guidance.

In any event, all these regulations--NYC100, CMS, and CDC-Prov--placed the burden for determining what was and was not essential on physicians and hospital administrators. NYC100: "Hospitals and amublatory surgery centers are directed to identify procedures that are deemed 'elective' by assessing which procedures can be postponed or cancelled based on patient risk considering the emergency need for redirection of resources to COVID-19 response"; CMS: "Decisions remain the responsibility of local healthcare delivery systems, including state and local health officials, and those clinicians who have direct responsibility for their patients." Or the NYT article linked above: "If physicians at EvergreenHealth believe a surgery should still occur, they could appeal to a panel [at the hospital]"--if this were a government mandate prohibiting specific procedures, this kind of self-regulation would not occur.

The government guidance and directives were not forbidding access to a specific list of procedures; they were instructing systems to triage: to reschedule some surgeries when, in the professional opinion of private experts, the procedures needed to be deprioritized. Any care postponed through this mechanism was not denied because of government action or policy; it was denied due to the coronavirus itself, to protect patients and providers from the risks of unnecessary exposure, and the anticipated need for resources to respond to the virus, as determined by non-government actors.

I will also note that the examples given in Atlas are mostly non-surgical in nature, and so would not have been affected by cancelling elective surgeries. I have already quoted NYC102, PAUSE, and the 202.6 Guidance calling out that providers of such care should continue. Moreover, CMS and CDC-Prov also indicate that the kinds of foregone care mentioned in Atlas should be provided. Take the category of childhood vaccinations. The CMS guidance categorizes these as "Tier 2, Intermediate acuity," and does not recommend postponing service. For CDC-Prov, for anything categorized as "Deferral of in-person care may result in patient harm" (pediatric vaccinations being specifically called out), the recommendation to providers is to "arrange for in-person care as soon as feasible with priority for at-risk populations"--even in areas with "substantial community transmission" of COVID. For areas with less COVID risk, providers are urged even more strongly to ensure that all needs are met. In the same document, "Signs/symptoms of stroke or heart attack" and "Treatment for certain cancer diagnoses" are listed as "Provide care without delay," even in environments with substantial community transmission.

Government policy, even during the lockdown, was to seek and to provide care for the types of conditions Atlas cites.

It can be argued that economic factors related to the shutdown might have restricted some people's access to preventive health care due to clinic closures from lost revenue; however, Atlas does not actually make this argument, which moreover would need to be quantified before it would be germane in an accounting-based analysis. Moreover, even if it were true, the obvious answer would be government support for health care facilities hurt by stopping elective procedures--the same way the government has in recent memory bailed out the investment banking industry in 2008-2009, the auto industry in 2008-2014, and the airline industry after September 11th. Such clinic closures support the claim that the government response was too little, as opposed to Atlas' main thesis that the government response was too much.

Point 2: Did government policy cause individuals to forego care due to fears that they would not have had in the absence of quarantine measures?

This is somewhat less definitive than point 1, but the claim that fear of COVID is caused by government policy--instead of, well, by COVID--is non-obvious and requires substantial more support than the mere throwaway assertion provided in Atlas. Ultimately, even careful surveying would not provide a definitive answer, because people are poor narrators of their own subconscious motivations and are prone to all kinds of hidden biases in thinking, as Atlas co-author John Hammond ought to know.

Nevertheless, there is plenty of evidence that fears of acquiring COVID in a healthcare setting are the rational response of people correctly estimating their own risks or choosing justifiable caution. One recent CDC study showed a drop of 42% in ER visits during the crisis. The reduction was largest among children (not medical decision-makers), women, and people in the Northeast--the latter group, of course, really did have the highest actual risk of infection during the period studied (by virtue of having the highest infection rates).

Moreover, there is epidemiological evidence that fears of infection in the health-care setting were justified: the first study I found showed COVID infection rates in-hospital at 25%, vs. 0.37% in the general public. Another report indicates hospital-acquired infection accounting for up to a fifth of COVID patients. Other estimates were even higher--this may be in China, with different protocols, but the disease is clearly very infectious, and can be acquired readily in hospital environments. Avoiding healthcare is not due to irrational fear.

Even if the fear were not justified, an argument that government mandates--and not COVID itself--have caused fear of COVID would also need to contend with the inundation of the public in media messages describing the dangers of COVID. We have all read reports of the disastrous outcomes in northern Italy, including reports of people trapped with the corpses of loved ones--well before US lockdown measures began. We heard stories of the virus' undetected spread. The US lockdowns were only just beginning when we heard about Italian funeral homes being overwhelmed.

That's if we weren't already paying attention to Wuhan citizens being dragged away by the COVID police or welded into their own apartments as their neighbors were recruited for a total-surveillance spying program and the providers in their health system were widely infected.

Or maybe we weren't looking overseas, but only to pre-lockdown reports of first responders with no equipment to prevent COVID spread, and journalists warning that it would only get worse. Maybe we heard about health care workers sleeping in their cars, for fear of bringing a deadly and highly infectious disease to their families--would you want to go get care from people who consider themselves so dangerous they isolate from their spouses and children? Maybe we listened to April-era reporting on health care workers improvising PPE from snorkels, even though those improvisers got infected anyway. Or maybe the people delaying care heard about how providers remained critical, well into the lockdowns, of a government response they described as inadequate.

Given the overall nature of reporting on the COVID situation in China, in Europe, and in the United States--and given that domestic reporting on COVID has been so critical of the inadequacy of American lockdown measures--claims that fears have been driven by overly strict lockdown itself are not credible.

Finally, declines in hospital admissions rates for non-COVID-related diseases are not a US-specific phenomenon. Two of the footnotes in CDC-Prov above document the same response among Northern Italian cardiac patients and Austrian cardiac patients. The article on the Italian case notes that this phenomenon was already statistically significant before Italian lockdown measures had gone into effect. Italian and Austrian patient behavior cannot logically be blamed on American quarantine policy; and if American patient behavior is in line with international trends, it is difficult to credit an argument blaming the trend on the American government's response.

Conclusion

Clearly healthcare delays are happening. But if the government did not shut down access to the kinds of health care described in Atlas--and if the government response cannot reasonably be blamed for people being shy of seeking out health care--what conclusion is left? Only that triage measures were due to coronavirus itself, that fear of coronavirus is justified by, and results from the risks of, coronavirus itself; in short, that these lost lives can only be blamed on the coronavirus itself.

Probably yes

The authors did a great job of citing sources and, except for underestimating COVID-19 years lost, the math largely checks out.

The obvious caveat is these figures are extrapolations; historical trends may not hold during the pandemic. For example, the correlation between unemployment and death may be weaker if more healthy, emotionally stable people are added to the unemployment ranks. Or perhaps the opposite: the correlation is even stronger since more households are losing all income sources.

There is also a question of how long conditions must last to actually produces their statistical effects. However, three months into quarantine with the end seemingly far off, that objection seems less significant.

Important: The quoted claim does not necessarily mean the net effect of quarantine is harmful. It simply compares negative effects of quarantine to negative effects of quarantined COVID-19. It does not consider lives saved by quarantine. That would be a separate claim and would require different analysis (which frankly, would be difficult, since it would be based on less common statistics...not to mention the unknowable possibility that quarantine might eventually fail). Despite not answering "the ultimate question" of the optimal quarantine level, the claim as quoted is scoped and its contents still offer relevant data points for that larger discussion.

Years lost due to unemployment deaths

One portion of this effect is through unemployment, which leads to an average increase in mortality of at least 60 percent. That translates into 7,200 lives lost per month among the 36 million newly unemployed Americans

Assuming these deaths occur proportionally across the ages of current U.S. mortality data, and equally among men and women, this amounts to more than 200,000 lost years of life for each month of the economic shutdown.

Unemployment

From mid-March to late May (when article was published), 39 million people in the US had filed for unemployment, though Forbes believed unemployment to be significantly higher than the number of claims.

Unemployment mortality

The article cites a 1994 study for the 60% mortality rate. A 1998 study finds a 30% increase in mortality (60% increase in accidental or violent death). A 2014 Scottish study found 85% and 50% increase for men and women respectively. A 2003 Swedish study found a 50% increase.

The estimates vary. However, even reducing 60% to 40% increase is insufficient to alter the conclusion.

Deaths

The U.S. mortality rate is 0.86% according to the CDC. For a population of 36 million, an additional 0.52% mortality would produce 187,000 additional deaths per year, or 15,600 per month.

The authors calculated half that many, probably by looking at the working-age mortality rate. Mortality scales approximately linearly from 0.07% for 15-24 to 0.89% for 55-64 (CDC). Assuming workers are distributed evenly across that, we approximately half the number, yielding a result only a little higher than the article's 7,200.

For workers aged 20 to 64, the lost years according to life expectancy is 20 to 59. For 7,200 deaths per month, that is 144,000 to 425,000 years lost per month. The authors stated an assumption of proportional spread, but even 5:1 weighting toward older workers still produces their 200,000 years per month.

Years lost due to reduced health care

Emergency stroke evaluations are down 40 percent. Of the 650,000 cancer patients receiving chemotherapy in the United States, an estimated half are missing their treatments. Of the 150,000 new cancer cases typically discovered each month in the U.S., most...are not being diagnosed, and two-thirds to three-fourths of routine cancer screenings are not happening...Nearly 85 percent fewer living-donor transplants are occurring...In addition, more than half of childhood vaccinations are not being performed, setting up the potential of a massive future health disaster.

The implications of treatment delays...result in 8,000 U.S. deaths per month of the shutdown, or about 120,000 years of remaining life. Missed strokes contribute an additional loss of 100,000 years of life for each month; late cancer diagnoses lose 250,000 years of remaining life for each month; missing living-donor transplants, another 5,000 years of life per month — and, if even 10 percent of vaccinations are not done, the result is an additional 24,000 years of life lost each month.

These unintended consequences of missed health care amount to more than 500,000 lost years of life per month, not including all the other known skipped care.

Health

It cites the following sources as evidence, and this gets pretty technical. I spot checked cancer, the largest source, and found agreement. If anyone looks at the others please share.

https://www.medscape.com/viewarticle/930374

https://www.cdc.gov/cancer/preventinfections/providers.htm

https://www.upi.com/Health_News/2020/04/01/Cancer-patient-care-disrupted-by-COVID-19-pandemic/7251585762174/

https://www.cdc.gov/cancer/dcpc/research/articles/cancer_2020.htm

https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045(20)30265-5/fulltext

https://www.docwirenews.com/docwire-pick/hem-onc-picks/report-medical-tests-including-cancer-screenings-have-rapidly-declined-since-covid-19-pandemic-hit-u-s https://unos.org/covid/

https://medicalxpress.com/news/2020-03-coronavirus-parents-kids-vaccinations-unicef.html

https://www.fraserinstitute.org/sites/default/files/effect-of-wait-times-on-mortality-in-canada.pdf

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5839705/

https://www.canaryfoundation.org/wp-content/uploads/EarlyDetectionFactSheet.pdf

https://unos.org/data/transplant-trends/

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5779444/

Years lost due to COVID

In comparison, COVID-19 fatalities have fallen disproportionately on the elderly, particularly in nursing homes, and those with co-morbidities. Based on the expected remaining lifetimes of these COVID-19 patients, and given that 40 percent of deaths are in nursing homes, the disease has been responsible for 800,000 lost years of life so far.

Death share by age

On June 19, the CDC reported the proportions of deaths by age. Combining this with life expediencies,

• <30 (51-79 years remaining) - 0%
• 30-39 (41-50 years remaining) - 1%
• 40-49 (32-41 years remaining) - 3%
• 50-59 (24-31 years remaining) - 8%
• 60-69 (16-23 years remaining) - 17%
• 70-79 (10-15 years remaining) - 25%
• >79 (0-10 years remaining) - 46%

So there are 9-16 years lost per death. Note this does not take into account pre-existing conditions which have disproportionately many COVID-19 deaths and also lower life expectancy (so the real number of years lost is likely lower).

Deaths

Through May 23, there had been 97,000 confirmed deaths in the U.S. from COVID, yielding 873,000-1,520,000 years lost. This is significantly more than the 800,000 that article stated.

I initially used life expectancy from birth (79) to do the calculation and had a number similar to theirs; perhaps they made the same mistake. Or perhaps they somehow accounted for the pre-existing conditions. Or used less recent death totals.

Conclusion

• Quarantine unemployment - Article: 200,000 yrs/month Check: 285,000 yrs/month
• Quarantine healthcare - Article: 500,000 yrs/month Check: 500,000 yrs/month
• Total quarantine over 2+ months - Article: 1,500,000 yrs/month Check: 1,680,000
• COVID-19 - Article: 800,000 yrs Check: 1,230,000 yrs

As the authors note, the quarantine deaths include only unemployment and healthcare; there are several other factors (e.g. reduced income but not unemployment causing substance abuse) that are not counted in that total that could bolster their argument.

In summary, the COVID-19 figure there is compelling evidence that there were more lost years from economic shutdown than from COVID-19, though the authors appear to underestimate COVID-19 years lost.

^{Note: You cited the article as claiming 700,000 lost years from COVID-19; it actually claimed there were 800,000.}

---

EDIT 1: Moved disclaimer about claim scope to top, as suggested by @IMSoP.

EDIT 2: Used actuarial tables for life expectancy, as suggested by @NuclearWang.