For other COVID posts, visit my Quarantine blog.
Throughout this COVID crisis federal and state governments have relied on academic models that were 100% wrong, 100% of the time. The media has willingly legitimized these marginal models because they hyper-sensationalized death and despair and that’s what sells newspapers and garners online viewership. The star students in the growing list of academics who tarnish their reputations and embarrass their Universities come from the University of Washington (UW). Their model has continuously been so outlandishly wrong I can never again read a UW paper believing it’s based on sound science.
The latest revision of the previously re-revised UW COVID model predicts that 243,000 Americans will die from COVID by August 4th. It’s never been clear to me if this prediction is based on actual COVID deaths or if it includes folks like that guy from Colorado who the coroner ruled died from alcohol poisoning but the CDC counted as COVID caused. I assume it’s based on CDC accounting, which means they project another 148,271 people die from something labeled COVID caused in the next 75 days. This equates to 2,004 people per day.
My issue with the UW model is that it was developed by MDs rather than PhDs and there’s a difference in expertise; particularly in understanding how to interpret scientific data. We now have a new COVID modeling team on campus from the University of Pennsylvania, Wharton School of Business. The ridiculousness of a COVID model emanating from a business school stands on its own, but what pushes it to the absurd is the mainstream media touting this model as legitimate because it aligns with their narrative.
The fine folks at Wharton have estimated, based on their years of scientific inquiry, that 117,000 people will die from May 1st to the end June if lockdowns continue, which translates to 2,187 people per day between now and the end of June. Their model further suggests if we end lockdowns, 350,000 people will die between May 1st and June 30th, which is 8,161 people per day between now and the end of June. To put these numbers in perspective, at the apex of the COVID crisis, 4,900 people died in one day and the average throughout this crisis is 1,128 people dying per day (of something the CDC claims is COVID caused).
This means the upper range of the Wharton projection is twice what the COVID crisis experienced at its peak and seven times higher than the average daily death rate. If we are to believe Wharton, May and June will be significantly worse than anything we’ve experienced so far and given we’re already mostly through a relatively mild May, they’re predicting June to be wildly catastrophic.
I question whether professors at Wharton or UW ever lecture on the need to validate models with either first principle calculations or actual data? Just because you successfully compile your software and get an answer when you executed the code, is not indicative of the answer being correct or even reasonable (ref. Dolin’s First Fundamental Law of Numerical Methods).
Let’s see if we can pinpoint more precisely how the Wharton model compares to the UW model. According the NY Times, half the country has reopened, and if we assume conservatively that the Wharton projection is linear, then their model predicts 201,794 Americans die from COVID between now and the end of June, which would make the cumulative death count at the end of June 296,523. Meanwhile, the UW model predicts a cumulative death count on that date of 172,872.
In order for the Wharton model to be valid, 5,174 Americans have to die each day between now and the end of June. In order for the UW model to be valid, 2,004 Americans have to die each day between now and the end of June. I do not believe either of these models will be validated because the CDC reported death rate has been trending downward since mid April (ref. left figure at top).
My model predicts 98,000 U.S. COVID deaths by June 30th, with the crisis over in early June. After that, the CDC reported deaths will be from flu. My prediction is significantly less than these academic models. But, my model assumes the CDC stops playing games with their death count reports. Maybe UW and Penn have insider information about how the CDC intends to count COVID deaths in June and their models are intended to shore up that preconceived narrative?
