U.S. Surpasses 150,000 Coronavirus Deaths, Far Eclipsing Projections

“Models are really only as good as the assumptions that you put into the model, but when you start to see real data, you can modify that model, and the real

data are telling us that it is highly likely that we’re having a definite positive effect,” Dr. Fauci said on the show, later adding: “It looks more like the 60,000 than the 100,000 to 200,000. But having said that, we’d better be careful that we don’t say, ‘OK, we’re doing so well, we can pull back.’”

By May 1, several states were reopening gyms, salons, restaurants and other businesses. Mr. Trump, who has given a wide range of predictions for the ultimate death count, said on May 3 that the virus might end up killing 100,000 people, after saying for much of April that the virus would not kill more than 75,000.

The next day, the Institute for Health Metrics and Evaluation, whose model is closely watched by the White House, increased its own projection, warning that there would likely be about 135,000 deaths by early August. The institute’s model, which includes a wide range of possible scenarios, now projects about 220,000 deaths by November.

Adding to the difficulty of predicting human behavior, Professor Pitzer said, is that public policy can be influenced by the models: seeing a forecast may prompt officials to take actions that make the forecast less likely to come true.

“Models are useful for playing out scenarios, but they’re not really meant to be accurate in generating long-term predictions,” she said. “They can be good at short-term forecasting — what might happen in the next couple of weeks. But longer term, knowing exactly what the trajectory of the epidemic will be —  there are just too many variables.”

Sarah Mervosh contributed reporting.