Read The New Yorker’s complete news coverage and analysis of the coronavirus pandemic.
I never found out how close Fermi got. What I took away from the story was a lesson about the power that numbers have to tame the natural world—to slice out the sizes and shapes of things. I’ve thought of that story throughout the coronavirus pandemic, when a few key figures seem to define our collective situation: the case count, a number that rises each day; the number of new cases, a picture of the pandemic’s trajectory; the number of confirmed deaths, a grim account of the virus’s toll.
Early last year, these numbers were in the headlines most days. With so much of ordinary life suspended, they acted as markers of pandemic time. On January 21, 2020, the first known case of the new coronavirus was confirmed in the United States; by the end of April, a hundred days later, more than a million Americans had been infected, and more had died from COVID-19 than were killed in the Vietnam War. At the beginning of the pandemic, these numbers held a foreboding significance. They told us where we were and suggested, unsettlingly, where we might go.
But now, a year later, the milestones fly by. At the height of this winter’s coronavirus wave, each week brought more than a million new infections in the United States. At least twenty-eight million Americans have now contracted the coronavirus—nearly a tenth of the country’s population. The virus has now killed half a million Americans, about a hundred thousand of them in January, the pandemic’s cruellest month so far. Such numbers blur in the mind. Stalin is said to have claimed that the death of one man is a tragedy, but the death of millions is a statistic. The idea resounds ominously today. Half a million Americans dead—a shocking number, at least until we reach the next one.
Look closer and the numbers themselves start to blur. The tally of confirmed coronavirus cases—now more than a hundred million worldwide—remains an underestimate. As the virus surges, testing capacity is regularly overwhelmed: in early January, test-positivity rates in several states reached record highs of between twenty and thirty per cent, suggesting that an enormous number of infections were still going undetected. And there is striking ambiguity about the seemingly straightforward number of COVID-19 deaths. Based on current records, more than 2.4 million people have died of COVID-19 around the world—but it’s clear that this official tally is far too low. According to an analysis of data from thirty-two countries and four major cities, six hundred thousand more people died in the first seven months of 2020 than in the same period in previous years; only about four hundred and thirteen thousand of these “excess deaths” were counted in COVID-19 mortality statistics. In the United States, excess deaths remain about twenty per cent higher than the official coronavirus death toll. They include the deaths of people who were never tested for the coronavirus, and non-COVID-19 deaths that might well have been avoided if not for the pandemic. This discrepancy suggests that our statistics continue to lag behind the reality of the moment; it may be years before we understand the true toll of the virus.
The number has lodged itself in my memory for several reasons—most of all, because of what scholars later called its “implausible degree of precision.” In high-school science class, we learn to think about numbers in terms of their “significant figures”—that is, their degree of precision, expressed through rounding. At one significant figure, Jordan’s estimate of the pandemic’s mortality would be rounded to twenty million. At two significant figures, it would be rounded to twenty-two million. From this perspective, there is something discomfiting about the number 21,642,283: its eight significant figures assert brazen certainty.
Jordan himself acknowledged some ambiguity: he wrote that the number of people who died during the 1918 pandemic was “probably at least as great” as his estimate of 21,642,283. He was right. Estimates of the death toll have continued to rise. A 1991 study that used updated records and statistical methods calculated thirty million deaths. The most recent and widely cited estimate, from 2002, found that fifty million people had died during the flu pandemic—although the authors write that the number may actually have been as high as a hundred million.
Fifty to a hundred million! These are staggering numbers, not just in their scale but also in the scale of their uncertainty. The world was at war during the 1918 flu pandemic, and some forty million soldiers and civilians died in those four years of global conflict. Still, their deaths fall within the pandemic study’s margin of error. I run over these estimates of the death toll again and again—21,642,283, thirty million, fifty to a hundred million—and the magnitude of the uncertainty unsettles me. These numbers make starkly different claims. Why is there so much disagreement? Why is the death toll of the 1918 pandemic so difficult to calculate?
First, there is the problem of missing records. In 1918, many countries had never conducted a modern census, and official registration systems, where they existed, were often disrupted by war. In colonial Asia and Africa, which were hit especially hard, records often overlooked the deaths of indigenous peoples. British India, which may have accounted for some forty per cent of flu deaths, is the largest source of uncertainty: initial British accounts estimated that six million people had died there, but in the early nineteen-twenties that number was revised upward, to twelve million, and then upward again, decades later, to seventeen and even twenty million—figures almost as high as Jordan’s initial global estimate of 21,642,283.
Even where mortality statistics exist, it can be hard to know what they mean. In 1918, the influenza virus had not been identified. There were no diagnostic tests, and causes of death were sometimes little more than guesswork. Some cases were so severe that doctors were slow to recognize them as influenza. Early in the pandemic, there were rumors that the disease was a return of the Black Death: critically ill influenza patients sometimes had such low oxygen levels that they started to turn blue; the color evoked the blackened skin and tissues characteristic of plague victims. Later, after influenza was established as the cause of the pandemic, some countries chose to include only influenza deaths in their official mortality statistics, leaving out the common and often deadly complication of bacterial pneumonia.
Then there were the pandemic’s long-term effects. Between 1917 and the late nineteen-twenties, about a million people worldwide came down with a mysterious sleeping sickness called encephalitis lethargica. Its cause remains unknown, and its symptoms, which ranged from insomnia to deep coma, varied so widely that its status as a single disease remains in doubt. But doctors at the time suspected links to influenza, and perhaps some of the half-million deaths attributed to encephalitis lethargica should be included in the pandemic’s toll as well.
These uncertainties are enough to drive a scientist to despair. The medical historians Niall Johnson and Juergen Mueller, in their 2002 global estimate of the death toll from the 1918 pandemic, quote a gloomy perspective on medical statistics from a British public-health report published in 1888. “It is useless . . . to shut our eyes to the imperfections of our records,” the report reads. “It is far better to be without statistics at all than to be misled by false ones.”
Still, even in this digital age there are limits to what we can know. Our most recent influenza pandemic, caused by the H1N1 flu, began in 2009; in its first year and a half, there were approximately eighteen thousand five hundred laboratory-confirmed deaths. But later estimates of global mortality, published in the years following, increased that number by fifteenfold, to more than two hundred thousand “respiratory” deaths, plus eighty thousand from cardiovascular complications. The reasons for the initial underestimate are familiar: incomplete testing, especially among those who died of complications; missing and incomplete records from poorer countries, where mortality rates are estimated to be two to four times higher than elsewhere.
In this pandemic, too, the limits of our knowledge are clear. Everywhere the virus surges, health-care systems are overwhelmed; hospitals scrambling to care for patients must sometimes set record-keeping aside. People start dying at home. Patchwork local and state regulations for certifying and reporting deaths mean that numbers are difficult to compare.
Because of these realities, the numbers are still changing. In April, after the virus’s spread in Wuhan was contained, city officials there revised their estimates of the local death toll upward by nearly fifty per cent, to account for probable COVID-19 deaths that were never officially confirmed. In many African countries, most deaths are still never officially registered, making it nearly impossible to understand the pandemic’s toll. And our definition of coronavirus-related deaths keeps expanding. Respiratory failure and pneumonia are the virus’s most common deadly effects, but the coronavirus can also cause heart attacks and severe strokes, complications that were often overlooked in the pandemic’s early days. A small number of COVID-19 patients go on to develop lasting “brain fog” or even severe psychotic symptoms. A chilling number of COVID-19 “long-haulers” experience symptoms for months, and, as the pandemic continues to unfold, we’re sure to learn more about these lasting effects.
In our polarized political climate, the uncertainty inherent in COVID-19 statistics becomes an opportunity for misinformation and disinformation. Right-wing commentators, urged on by former President Donald Trump, frequently allege that the death toll is inflated, even though the number of excess deaths suggests the opposite. In August, conspiracy theorists seized on cause-of-death statistics from the C.D.C., claiming that only nine thousand Americans had died of COVID-19—just six per cent of the official toll at the time.
What the statistics actually showed was that only six per cent of U.S. COVID-19 deaths were attributed to the disease alone: the other ninety-four per cent involved underlying conditions such as asthma, diabetes, or heart disease. On death certificates, coroners specify a chain of events, a domino effect that ends in death. An “underlying cause” of disease, such as COVID-19, sets the chain in motion. The initial infection can progress to pneumonia, which can, in turn, become respiratory failure—the final domino that is the “immediate cause” of death. To complicate things further, there are also “contributing factors,” such as obesity or diabetes, that influence the disease’s progression—but here the lines of cause and effect multiply into a web. These contributing factors do not lessen the virus’s impact; if anything, they have multiplied its toll.
The death toll will continue to rise long after the pandemic ends. Thinking of this, I remind myself of an offhand comment that has lingered in my mind—an outlook expressed, in the nineteen-eighties, by a group of biostatisticians grappling with the challenge of knowing anything in an unknowable world. “We see our business as one of lighting candles rather than cursing the darkness,” they wrote. To understand the full devastation that the virus has wrought, we’ll need many more candles to light the way.