Revisting UW IHME’s Covid-19 projections

This part of the post was written on September 4th, but not published until December 30.

On September 4th, the UW IHME disease modeling team projected 410,000 deaths in the U.S. by January 1st, 2021:

The Institute for Health Metrics and Evaluation at the University of Washington’s School of Medicine is predicting more than 410,000 deaths by January if mask usage stays at current rates. If governments continue relaxing social distancing requirements, that number could increase.

Source: Coronavirus live updates: Model predicts 410K US deaths by January; Labor Day weekend brings risk; South Dakota stages state fair

By September/October, most of the US was averaging 90+% face mask compliance. The areas with bad outbreaks were seeing 95+% compliance according to COVIDcast at CMU.

On September 4th, the CDC reported a cumulative total of just over 187,000 deaths.

As of December 31st, the CDC reports:

Including Dec 31 in the final tally will likely bring that to over 340,000 for the year.

Since last April, I saved many of the UW IHME projections. Their projections, when compared to reality, ranged from being wildly off to off in space.

In every projection I saved, the UW IHME significantly over estimated future deaths.

As they got closer to December 31st, IHME revised their past projection to 296,000 to 348,000 – moving their own goal posts so they could later say, see, we made a correct projection! And they can claim with a straight, but covered, face that the real world was at the high end of their projection!

UW IHME continuously updates their past web site projections, much like the Soviets used to rewrite history – so that at any point in time, their projections look really good 🙂

They also insert a random variable of unknown value – face mask usage – such that they can later justify any possible real world scenario. The CDC itself has said it has no quantitative data to compare face masks versus social distancing or hand washing or any other mitigation. In other words, UW makes up assumed values for their face mask wearing model – meaning they can generate any projection and then later blame errors on face masks. IHME assumed – back in May – that wearing a mask provided a 50% risk reduction. This was based, then, on early comparisons that ignored the time dimension. For example, CA “proved” that mitigations and mask wearing worked – In May! Now in Dec, CA has the worst outbreak nationwide after having the toughest restrictions for the longest period of time compared to elsewhere. It looks like IHME is making up a fudge factor value so their projection can always agree with the narrative du jour.

Their projections have been consistently awful. Yet much of the media – whose business uses gloom and doom click bait to sell eyeballs to advertisers – eagerly publicize these “expert predictions” as if they were meaningful.