A common talking point in response to the periodic updates to model projections is that “the models have all been wrong”. This stems in part from Donald Trump’s repetition of the talking point that the models predicted “millions of deaths”.

It is true that, VERY early on, models based on assumptions that we did NOTHING – no distancing, no shudowns, no masks – predicted millions of deaths. In fact, we did a lot so we had many fewer deaths than that. But even so, I don’t think that prediction is unreasonable given that we may be 1/3 through the crisis and already have almost 200,000 deaths with all of our mitigation. But it’s hard to know for sure.

What we do have is an actual record of projections and how things worked out. I’ll work with the most popularly cited model, the IHME. They post their prediction updates on their press page, so we can track them.

3/2681,000 deaths by July 26147,000
4/4135,000 by early august155,000
6/11169,890 by October 1198,000+
6/24180,000 by October 1198,000+
7/1200,000 by November 1198,000+

What we can see is that the IHME model has a track record of under-predicting deaths. So when it currently predicts 415,000 deaths by Jan 1, history would imply it will actually be quite a bit more.

In fact, we can see that this prediction is reasonable. Even if you only believe the current rate sustains, that’s going to be around 100,000 more cases for a total of around 300,000. IHME is predicting we will have 215k more deaths in 110 days. This 2000 deaths per day number is an acceleration from the current rate of around 900 per day, and clearly assumes that as winter comes, people will be inside more and there will be an acceleration as is typical of Coronaviruses.

Is that assumption right? We’ll see. But given their track record, I wouldn’t bet against it.