Microbiologist
© David Ryder/ReutersMicrobiologist Milagros Sola processes coronavirus tests in a lab at Madigan Army Medical Center at Joint Base Lewis-McChord, Wash., April 14, 2020.
One of the biggest unknowns regarding COVID-19 is how many people get the disease, recover from it, and develop immunity without being captured in our official statistics — perhaps without showing symptoms at all. Some suggest this number could be incredibly high indeed, implying that the disease spreads quickly but has a lower-than-expected fatality rate.

We should have good answers on this soon thanks to "serology" tests, which detect whether someone has antibodies for COVID (meaning they had the virus at some point in the past even if they are not currently infected). But here I'd like to offer a quick overview of what we already know. My own guess is that while we are substantially undercounting cases, it is unlikely that more than, say, 10 percent of the population already has it. Since about two-thirds of the population needs to get it to achieve "herd immunity," that would give us a long way to go if we just lift the lockdowns and hope for the best.


Comment: That's assuming lockdowns even work, and that the costs they incur (monetary AND human) don't overtake the costs of just "hoping for the best". It may be that in the long term, this is just another virus humanity has to pass through. As it is, the mortality rate either similar to, or perhaps 2x, that of the ordinary flu. It's no black death, and certainly won't end up decimating the population. The containment measures however, have a higher chance of doing just that.


To begin, some very simple numbers. At this writing, the U.S. has about 600,000 confirmed cases and a population of about 330 million people. If we are undercounting cases by a factor of ten, that still puts us around 2 percent. To get to 10 percent infected we'd need to be undercounting by a factor of about 50.

So what does the evidence say? Well, there's a lot of it, pointing in a lot of different directions. Here's a sampling:
  • To start with the most optimistic piece, a study recently profiled in The Economist points out that states with high rates of COVID-19 also had a lot of amorphous "influenza-like illnesses" earlier in the year. If these illnesses were actually undetected COVID, that would mean lots of cases with symptoms weren't counted. All told, we could be undercounting cases by a factor of 100 or even 1,000.
. . . well, actually, probably not 1,000, since that would give us more infections than people. Even 100 is officially impossible in New York State now, because it's diagnosed 1 percent of its entire population.

At the national level, 100-fold undercounting would imply that 20 percent of people have already had it. Cross your fingers but don't bet on it. On a side note, I recommend this thread from the author of the Economist piece, which provides a lot of context the story doesn't and forcefully argues against lifting containment measures. (It also links a bunch of data not included in this piece, for those who want a deeper dive.)
  • As I've noted on the Corner, a New York hospital tested women giving birth and found that 15 percent were currently infected, overwhelmingly without symptoms. Of course, however, New York is the hardest-hit place in the country, while childbearing-age women are far from the population most susceptible to severe symptoms, making the numbers hard to generalize.
  • San Miguel County in Colorado, a state not particularly affected, is doing antibody tests. Early results for nearly 2,000 people identified just 11 positives and 30 borderline cases. Gangelt, Germany, which was hit hard, found antibodies in 15 percent of its sample (suggesting a fatality rate of 0.4 percent, which would kill about 800,000 Americans if two-thirds of us got it in the pursuit of herd immunity). Iceland, looking for current infections in a random sample of its population, found them less than 1 percent of the time, and about half of those who tested positive reported symptoms. Another paper from Iceland estimates the overall undercount at ten times.
  • Relying on a combination of the Iceland data and some U.S. travel information, a recent study estimates we're catching somewhere between 1.5 percent and 14 percent of infections. That's an undercount of seven- to 67-fold. Another paper says the U.S. government was catching about 7 percent of infections around late March.
  • A Boston homeless shelter found a current-infection rate above one-third in a sample of a few hundred, overwhelmingly asymptomatic. (Indeed, those who tested positive showed about the same level of coughing, fevers, etc., as those who tested negative.) However, viruses will obviously spread quickly in a homeless shelter, and some of the sample developed symptoms after the data were collected.
  • There's a ton of variation in mortality across countries and U.S. states. Some of this might be chalked up to differences in reporting, not to mention demographics and the time gap between infection and death, but it also likely suggests that "lagging" areas still have a lot of uninfected people. (In other words, even if the hardest-hit places already have herd immunity — a dubious assumption to start with — a lot of people will die if the rest are, er, "brought up to speed.") There are six countries with per capita death rates more than twice that of the U.S., and New York has about twice the death rate or more, usually much more, of every other state.
Make no mistake, these pieces of information all suggest we're undercounting infections by quite a bit. A ten- or even 20-fold undercount wouldn't shock me at all. But it strikes me as wishful thinking to believe the U.S. is anywhere near herd immunity already, which would require us to be undercounting by several-hundred-fold.