Good Reasons to Doubt the Estimate of COVID-19 Deaths

Many popular accounts mistake an implausible upper-bound estimate for the most likely outcome. The underlying study actually suggests the number of deaths will be much lower.

A recent study from the Imperial College COVID-19 Response Team estimates that as many as 2.2 million Americans could die from the coronavirus (COVID-19). Its estimates come from an epidemiological model which, among other things, takes into account the strain on hospitals that is expected if we are unable to flatten the curve

The New York Times credits the study with changing the tone at the White House, which revised its guidance on gathering limits from 50 to 10 on Sunday and urged Americans to increase social distance.

That headline number—2.2 million deaths in the US—has received a lot of attention. As the authors of the study note, however, that estimate only results in “the (unlikely) absence of any control measures or spontaneous changes in individual behavior.” In other words, it is the study’s worst-case scenario, where everyone just goes on as if there were no virus even though people around them are getting sick and dying. 

Perhaps the authors should have written “incredibly unlikely” and not put it in parentheticals. Again, the estimate is based on the implausible assumption that we do nothing. And, by “do nothing,” I do not mean “fail to adopt a meaningful government response.” There would need to be “no spontaneous changes in individual behavior” (i.e., those not directed by the government) as well.

Such an estimate might provide a useful starting point. In particular, it might give us some perspective on the absolute upper bound of reasonable estimates. But it should not be taken as a reasonable expectation of what will actually happen. Remember the Lucas critique: when circumstances change, people change their behavior. The headline estimate merely recognizes how bad it would be in the absence of any such behavioral or policy changes.

I cannot overemphasize how implausible the headline estimate is. We cannot “do nothing” at this point because we are already doing more than nothing. We are isolating those infected, banning travel from high-risk countries, self-quarantining the at-risk, working remotely, closing schools, increasing social distance, washing hands more frequently, … And those control measures and behavioral responses are precisely the kind of steps the study’s authors go on to consider.

If anything, I would say the study gives us some reasons to be optimistic. The authors estimate a 15 to 30 percent reduction in deaths from just two, small changes:

  • Case isolation in the home
  • Social distancing of entire population

Moreover, the authors are relatively conservative in modeling these changes. For example, they assume isolation reduces contact outside the home by just 75 percent and that only 70 percent of those experiencing symptoms actually respond by isolating. 

They assume social distancing reduces contact outside the home, school, or workplace by 75 percent, but that school contact rates are unchanged and workplace contact rates are reduced by just 25 percent.

The authors estimate a 49 to 50 percent reduction in deaths from three changes:

  • Case isolation in the home
  • Voluntary home quarantine
  • Social distancing of those over 70 years of age

Voluntary home quarantine differs from case isolation in that all household members stay home following the identification of symptoms, not just the person (or, case) showing symptoms. And, as in the previous scenario, the assumptions concerning compliance are conservative. Only 50 percent of people are assumed to voluntarily quarantine when one of their family members becomes ill. 

More strikingly, in this scenario, those of us under age 70 without symptoms or family members with symptoms are assumed to go on about our lives as if there isn’t a virus killing a bunch of people. We do not work remotely. We do not cancel upcoming trips. We do not increase social distance. We do not wash our hands more frequently. There is no behavioral response. It is just business as usual for most of us. That strikes me as implausible. And, the more we change our behaviors in response, the fewer deaths will result.

To recap, with three not-so-incredible responses, the authors of the Imperial College study estimate that we will reduce deaths from 2.2 million to 1.1 million. That is still a lot of deaths, to be sure. It is roughly 32 times as many deaths as resulted from the flu last year. 

However, we should keep in mind that the estimates produced in the study are based on relatively conservative assumptions about our responses and how the disease will spread in the US. I have written a lot about the former already, so let me now briefly consider the latter.

The authors implicitly assume that COVID-19 will move from person to person in the US just as it did in China and South Korea. However, we have learned that proximity matters a lot with this disease. A recent report from the World Health Organization (WHO) found that most human-to-human transmission of COVID-19 in China occurred in families.

Part of the reason why diseases spread so rapidly in places like China and South Korea is because there are so many people living so closely together. Despite vast rural areas, population density in China is still roughly 375 people per square mile. In Wuhan, where COVID-19 broke out, it is around 3,379 people per square mile. In South Korea, there are some 1,302 people per sq mile. And, in Seoul, its capital, there are 41,655 people per sq mile.

The US is much less densely populated than China and South Korea and Americans are much more likely to live alone. In the US, there are just 90 people per square mile. We have high-density areas, like New York City, where there are roughly 27,751 people per square mile. But most cities in the US are more like Columbus, OH—3,960 people per square mile. And most places in the US are not cities. That means, at least outside of a few large cities, COVID-19 will have fewer direct, personal contact points to spread from person to person in the US.

Based on my reading of the Imperial College COVID-19 Response Team study, I conclude that 1.1 million is a plausible high-end estimate of the number of deaths in the event that we take no extreme measures and only partially comply with sensible measures. In addition to the estimate, however, there are three key takeaways. 

First, none of the responses considered above requires government action. We can choose to isolate when we experience symptoms. The family members of those who fall ill can choose to quarantine themselves. We can choose to increase social distance. We do not need the government to force us to do those things.

It is also not merely a matter of individual choice, though. There is a big role for civil and commercial society to play. We can advise others to wash their hands more frequently; to make one trip to the grocery store per week rather than three; and to stand further apart when we talk. We can shun and shame those who refuse to heed our advice. We can postpone unnecessary social gatherings, or move them online. We can come up with novel business solutions, like carving out special times for those most at-risk to shop. We can share best practices for working remotely and provide resources to others in our field who are transitioning away from face-to-face encounters.

Consider some further examples from my personal experience. My employer has required any employee who has recently traveled abroad to self-quarantine for 14 days. Will I be shot or sent to prison if I refuse? No. But, in a state with at-will employment, I could lose my job. At the very least, I would receive some disappointing glances and perhaps justifiably unkind words from colleagues for putting them all at risk. So I am working from home, waiting it out.

My gym, which is usually packed and sweaty for a few classes each day, has capped the number of participants at 12 per class and significantly increased the number of classes offered to accommodate. For folks hunkered down like me, they have rolled out a separate program that can be completed at home with no equipment and little space. No doubt these efforts were difficult to develop quickly and more costly to provide than the usual services. But the owners care about the community they have built. And they are led by the profit motive to provide the services their customers demand.

Can government policies limit the spread of COVID-19 further still? Sure. The government has one core competency: it can use force. And, in most cases, it can force us to do more of something than we would do on our own. But we should be hesitant to permit use of that force.

As a liberal, I strongly believe in the presumption of liberty. That does not mean it is never acceptable to coerce others. Rather, it means that it is only acceptable to coerce others when there is a very good reason for doing so.

With a presumption of liberty, each person is mostly free to choose the extent to which they interact with others or engage in isolation. However, if one tests positive for or is exhibiting obvious symptoms of COVID-19, then it is justifiable to impose isolation on that individual and those they have closely interacted with for a reasonable period of time, under humane conditions. In cases where it is less clear that coercion is warranted, however, we should err on the side of liberty

The second key takeaway is that the relevant trade-offs depend crucially on local conditions. Outside of a few high-population-density urban areas, the cost of shutting everything down—in terms of real hardships for real people—is almost certainly unwarranted. We do not want the virus to spread. But we also do not want a cure that is worse than the disease.

Furthermore, the second takeaway suggests we should be looking for more decentralized solutions. The most appropriate policy response will probably vary from place to place. COVID-19 poses less risk to Omaha (3,378 people per square mile) than Chicago (11,841 per square mile). And it poses even less risk to rural areas. In Scioto County, OH, where I grew up, there are just 130 people per square mile. A one-size-fits-all approach will almost certainly result in far more costs than are warranted.

There are certainly things that federal and state governments can do to reduce the spread of COVID-19 and mitigate the economic burden for the least well off. They can collect and provide information, especially insofar as it relates to issues that cross jurisdictions. They can trace the close connections of those infected and support isolation efforts when warranted. They can delay when tax payments are due. But much of what needs to be done can—and should—be done locally.

The third key takeaway is that small behavioral changes can really add up. Recall that the study made relatively conservative assumptions about compliance and, yet, resulted in significant reductions in the total number of deaths in the US. It follows that we can probably limit the number of deaths even further than those estimated in the study by increasing compliance with sensible measures. If you can work remotely, you probably should work remotely. Instead of dining at a restaurant, order take-out. And wash your hands every 90 minutes, setting a timer if you are prone to forget. These small changes impose small costs. But they seem to yield outsized benefits.

To many, the Imperial College COVID-19 Response Team study is taken to mean the end of the world is nigh. Having read the report, I offer a very different conclusion. It suggests small, personal sacrifices are warranted; local businesses must come up with innovative solutions; communities should postpone large social gatherings until the storm has passed; and governments ought to trace the close connections of those infected and require isolation when warranted. The study is not the last word on the subject. Indeed, other scientists are already weighing in, discussing its strengths and weaknesses. But it is a sober assessment of the most pressing problem facing the world right now. And we should share and discuss it with the same level of sobriety.