1/ Let's get back to the (new!) @IHME_UW #COVID model - this time looking at states. Recall, the model cut its estimate of peak hospitalizations almost in half last night, but barely reduced the US death forecast, which remains at 81,000 (somewhat worse than a bad flu year)...

2/ Obviously, to get that estimate, the model adds state death totals. But some big states have VERY low estimates. California and Texas, have a total of about 70MM people, have under 4,000 deaths combined. PA has 782. (You can see all this yourself here: https://covid19.healthdata.org/projections 

3/ Okay, where are the deaths coming from? NY and NJ account for about 25,000, which makes sense given that the epidemic appears centered in NYC: but then a bunch of apparently random states. Massachusetts, for example, is projected to have 8,200 deaths; Florida, 6,800...

4/ The problem is that the CURRENT hospitalization trends in the "bad" states (again, aside from NY/NJ) are indistinguishable from the trends in the "good" states. Florida's new hospitalizations are dropping - but so are Massachusetts's (153 to 102 to 77 Friday thru Sunday)...

5/ So it is not quite clear why these states, which obviously are not part of the NY metro area - would suddenly follow a more New York-like path, when they haven't already...

6/ Which is just another reason to question whether the model at the most basic level is making any sense. Not that you needed one. The gold standard! (Which, ironically, we quit using the last time unemployment was this high.)