Nate Silver (@NateSilver538 )

Nate Silver

Bio Editor-in-Chief, @FiveThirtyEight. Author, The Signal and the Noise (amzn.to/QdyFYV). Sports/politics/food geek.
Location New York
Tweets 29,9K
Followers 3,3M
Following 1,2K
Account created 27-08-2008 20:56:45
ID 16017475

Twitter Web App : Yeah, the "leaked White House projections" showing 3K deaths and 200K! cases per day by the end of May 1) never made much sense 2) don't actually seem to have reflected the White House's view. The way the story was reported was a big mess from the get-go.
twitter.com/NateSilver538/… twitter.com/RAVerBruggen/s…

Twitter Web App : Accusing your opponent of stealing the election doesn't seem like a great way to increase turnout among your base.

Twitter Web App : More context on Georgia. Actually, 78K antibody tests, which they still include in their totals but which you can easily back out of the numbers since they now list them separately, as it looks like The COVID Tracking Project is doing.

twitter.com/Crimealytics/s…

Twitter Web App : The 7-day average death toll has fallen below 1,000 for the first time since April 2. The 7-day average case count has also ticked down slightly after plateauing last week. The weekend may still be skewing the numbers a bit, though, so possible we'll get bad news tomorrow/Friday.

Twitter Web App : The other thing is that I thought we'd get some states reporting large numbers of deaths and cases because of reporting lags after the 3-day weekend. But, actually not *that* bad. While 1,259 deaths is way up from the weekend, it's down from 1,402 last Wednesday.

Twitter Web App : * Test numbers *dropped* by ~74,000 in Georgia, which I assume means they removed antibody tests from their testing numbers (good!). Without GA, there were 360K tests reported, a fairly typical number for recent days, for a positive rate of 5%.

Twitter Web App : US numbers via The COVID Tracking Project. A weird one.

Newly reported deaths
Today: 1,259
Yesterday: 629
One week ago (5/27): 1,402

Newly reported cases
T: 19K
Y: 16K
5/27: 21K

Newly reported tests
T: 286K*
Y: 302K
5/27: 405K

Positive test rate
T: 7%*
Y: 5%
5/27: 5%

* See next tweet

Twitter Web App : Matthew Yglesias Conversely, if you take a pessimistic view, the implication could be "we need to go back to harsher lockdowns now" or it could be "this is hopeless so let's do Sweden".

Twitter Web App : Matthew Yglesias Ehhh, I think they're relatively orthogonal. Like, if you take a relatively optimistic view on current trends, the implication of that could be "great, our strategy is working, let's keep it up!" *or* it could be "OK, we can afford to loosen our belts a little bit".

Twitter Web App : It's true that the COVID data is a mess and it takes some work to figure out how to use it robustly. But people have to be careful not to attribute outcomes to bad/untrustworthy data every time their priors are disconfirmed.

Twitter Web App : Which is not to say you can't find fault with the critics. Trying to convince people that "reopening will cause huge spikes", when in practice re-opening was extremely modest and tentative, was a dumb and foreseeable mistake and set a low bar for success for these states.

Twitter Web App : Now that restaurants in these places are operating at ~50% capacity (not that restaurants themselves are so important, but using them as a proxy for broader activity) we'll see what the numbers look like in a couple of weeks.

Twitter Web App : One way to look at this: until the past couple of weeks, sit-down restaurants in TX/GA/FL were operating at maybe 10-20% capacity and returning only 10-20% back to normal (with people being careful, using socially distanced seating, etc.) was never likely to cause a huge spike. twitter.com/tomgara/status…

Twitter Web App : In other words, the flip side to not getting too excited about these numbers is not freaking out too much if states report big one-day totals tomorrow or Thursday, which is likely to happen in at least some places if their reporting is backlogged.

Twitter Web App : This is the fewest reported cases in a day since 3/25, but that's probably at least in part because states are slow to report at the end of a 3-day weekend, which likely means the numbers will bounce back tomorrow and Thursday.

Twitter Web App : US #'s via The COVID Tracking Project

**Likely impacted by holiday weekend**

Newly reported deaths
Today: 629
Yesterday: 523
One week ago (5/19): 1,430

Newly reported cases
T: 16K
Y: 19K
5/19: 21K

Newly reported tests
T: 302K
Y: 441K
5/19: 397K

Positive test rate
T: 5%
Y: 4%
5/19: 5%

Twitter Web App : Two interesting papers.

The claim of the latter paper—that pessimism about COVID-19 begets fatalism that reduces compliance with social distancing—is an experimental rather than empirical result. I'd be more persuaded by empirical evidence, but it's an interesting conjecture. twitter.com/DKThomp/status…

Twitter Web App : As states reopen, not enough talk about early indicators for detecting where there may be an uptick in spread. Anything from sewer data to thermometer data to Google search data for symptoms could potentially be helpful and quicker than waiting for test results. twitter.com/JonWalkerDC/st…