Instead, it had relied on a mix of data sent in by member countries and by academic groups, and on a meta-analysis done in May by scientists at the University of Wollongong and James Cook University in Australia.
Those researchers looked at 267 studies in more than a dozen countries, and then chose the 25 they considered the most accurate, weighting them for accuracy and averaged the data. They concluded that the global I.F.R. was 0.64 percent.
The C.D.C.’s estimate for the United States is lower: an I.F.R of 0.4 percent, according to a set of planning scenarios released in late May. The agency did not respond to requests to explain how it arrived at that figure, or why it was so much lower than the W.H.O.’s estimate. By comparison, 0.4 percent of the United States population is 1.3 million people.
The 25 studies that the Australian researchers considered the most accurate relied on very different methodologies. One report, for example, was based on diagnostic PCR tests of all passengers and crew aboard the Diamond Princess, the cruise ship that docked in Japan after it was overcome by the coronavirus. Another study drew data from an antibody survey of 38,000 Spaniards, while another included only 1,104 Swedes.
The current W.H.O. estimate is based on later, larger studies of how many people have antibodies in their blood; future studies may further refine the figure, Dr. Swaminathan said.
But there is “a lot of uncertainty” about how many silent and untested carriers there are, Dr. Morgan of the W.H.O. said.
To arrive at the C.D.C.’s new estimate, researchers tested samples from 11,933 people for antibodies to the coronavirus in six regions in the United States. New York City reported 53,803 cases by April 1, but the actual number of infections was 12 times higher — nearly 642,000, the agency estimated.