Without "much more aggressive shutdowns," a New York Times editorial warns, "well over a million" Americans "may ultimately die" from COVID-19. The paper does not cite a source for that estimate, which seems highly implausible based on the death toll so far, projections for the next few months, the gap between total infections and confirmed cases, and a crude case fatality rate that continues to fall.
Independent data scientist Youyang Gu, who has a good track record of predicting COVID-19 fatalities, is currently projecting about 231,000 deaths in the United States by November 1. The University of Washington's Institute for Health Metrics and Evaluation projects 295,000 deaths by December 1. Assuming those projections prove to be about right, the Times is predicting that the death toll will quadruple during the months before an effective vaccine can be deployed, which might happen early next year.
If we have a vaccine by April, for instance, the Times projection would require that deaths rise at a monthly rate of about 40 percent in December, January, February, and March. By comparison, according to Gu's figures, the death toll rose by about 19 percent in June and 21 percent in July. His projections indicate increases of about 21 percent this month, 13 percent next month, and 9 percent in October.
The expected tapering of the upward trend in total deaths may seem surprising in light of this summer's surge in confirmed cases. But according to Worldometer's tally, which is based on numbers from local and state health departments, the nationwide seven-day average of daily new cases, after rising above 69,000 on July 25, had fallen to less than 55,000 as of yesterday. And even as daily new cases rose dramatically in June and July, the nationwide crude case fatality rate—reported deaths as a share of known cases—continued to fall. The current rate, 3.2 percent, is about half the rate on May 15. There are several plausible explanations for that downward trend, including expanded testing, better treatments, and a younger, healthier mix of patients. But the decline in the fatality rate makes the death toll predicted by the Times look even less likely.
According to Worldometer's count, the current COVID-19 death toll in the United States is about 166,000, while the number of known cases is 5.2 million. But as the Times has noted, antibody research by the Centers for Disease Control and Prevention (CDC) suggests the actual number of infections may be more like 52 million, which indicates an infection fatality rate around 0.3 percent. Assuming those estimates are in the right ballpark, a death toll of "well over a million" would imply more infections than people.
Although the threshold for COVID-19 herd immunity is a matter of much dispute, many epidemiologists think 60 percent is a reasonable estimate. If so, the maximum number of U.S. infections, even without any precautions, would be less than 200 million. That means the Times is assuming a nationwide infection fatality rate close to the CDC's current "best estimate" of 0.65 percent.
Even if the nationwide IFR turns out to be that high, the death toll imagined by the Times is feasible only if all the precautions we have taken to curtail the spread of COVID-19—including legal restrictions on social and economic activity as well as voluntary measures such as physical distancing, avoiding crowds, mask wearing, and hand washing—prove to be ineffectual. In other words, the Times seems to be comparing its policy prescriptions to a counterfactual "no intervention" scenario, which does not bear even a remote resemblance to the choices we are actually confronting.