Nate Silver

Nate Silver @NateSilver538

* Obama is popular so going hard on Obamagate is a strange choice. Maybe Trump thinks these things are helping him. But if he were more rational about optimizing his re-election chances, his COVID response would likely have been a lot better. — PolitiTweet.org

Nate Silver @NateSilver538

* Trump's approval ratings for his handling of COVID are poor, and worsening. People notice the high death tolls. * Fast re-openings likely won't help the economy by Nov. if they cause an additional wave of disease, etc. * No evidence that mail-in voting helps Democrats. — PolitiTweet.org

Nate Silver @NateSilver538

I don't know what motivates Trump. But, it's unlikely these things are helping him re-election. It's more likely that they're hurting him. Consider: * The public favors maintaining lockdowns or slow, cautious re-openings. — PolitiTweet.org

Nate Silver @NateSilver538

This timing is interesting insofar as the US data started to look like it was getting a little better as of early May, at least to my eye. And if you had a big inflection point in mask-wearing in mid-April, that's when you might expect to see the effects. — PolitiTweet.org

Nate Silver @NateSilver538

@youyanggu @CT_Bergstrom Meaning like, 30% in Stockholm *city* as of 3 weeks ago? Nope, probably not. But 20-25% *now*? Maybe. https://t.co/jZpnej66NN — PolitiTweet.org

Nate Silver @NateSilver538

@youyanggu @CT_Bergstrom Cool, thank you! And the 7% rate in the study released today was for Stockholm County, not Stockholm proper? (Or do we know?) My priors (partly based on your model) had been that Swedes were being maybe 50% too optimistic in the estimates they were giving to the press. — PolitiTweet.org

Nate Silver @NateSilver538

@youyanggu @CT_Bergstrom If you buy their claim that it reflects the state of play as of early/mid April, what does your model say about the total number infected as of now in Stockholm? — PolitiTweet.org

Nate Silver @NateSilver538

@chrislhayes @AlecMacGillis @SeanTrende It's normal for people to be scared/careful. It is also asking a lot of people to synthesize a lot of rapidly-evolving news seems to be in conflict. At the same time, I think there's a gap between what people think "the science says" and what the science actually says. — PolitiTweet.org

Nate Silver @NateSilver538

FWIW, the numbers are coming in close to what the consensus of models projected a few weeks ago. (These are forecasts as of May 1.) So I'm not sure that the numbers are actually better than experts expected, so much as that there's a perception issue because of media framing. https://t.co/Y8ckCJ5RrN — PolitiTweet.org

Nate Silver @NateSilver538

FWIW, I was not a big fan of that media narrative. I thought it didn't do a great job of reflecting the actual range of expert opinion, the complicated, mixed signals in the data, or the fairly wide uncertainties in the outlook. https://t.co/BMgDwOb5jQ — PolitiTweet.org

Nate Silver @NateSilver538

These numbers are good, but not great. The improvement is steady, but slow. They could take a turn for the worse later. But at this point, folks have to contend w/ the fact that the media narrative was very pessimistic a few weeks ago. That will affect how they're perceived. — PolitiTweet.org

Nate Silver @NateSilver538

And there continues to be slow, but meaningful improvement in the fatality numbers. An average of 1,304 deaths in the past 7 days, versus 1,584 in the week before that. The 7-day average peaked (so far; keep in mind there could be a second peak) at 2,070 in mid-April. — PolitiTweet.org

Nate Silver @NateSilver538

Actually, 414K tests, not 411K. (Typo on my part.) This was a record day for tests! — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking: Newly reported deaths Today: 1,402 Yesterday: 1,430 One week ago (5/13): 1,693 Newly reported cases T: 22K Y: 21K 5/13: 20K Newly reported tests T: 411K Y: 399K 5/13: 326K Positive test rate T: 5% Y: 5% 5/13: 6% — PolitiTweet.org

Nate Silver @NateSilver538

The people who are confident are probably right, by the way. The media narrative hasn't caught up to the more widespread availability of testing, and it's probably deterring some people from getting tested when they could be. https://t.co/EVCQSX7N9x — PolitiTweet.org

Nate Silver @NateSilver538

This is pretty interesting. 85% of Republicans but only 51% of Democrats are very confident they could get a coronavirus test if they wanted one. https://t.co/DziRa8lZRW https://t.co/6py6X6WIEC — PolitiTweet.org

Nate Silver @NateSilver538

That is, you probably want a policy that is: * Nimble. You need to collect data quickly and be ready to adapt to it. * Multifaceted. There inevitably has to be some element of "trial and error" as you learn about what works and what doesn't. — PolitiTweet.org

Nate Silver @NateSilver538

In general, critiques of governmental responses to coronavirus don't do enough to account for the fact that there are a lot of unknowns surrounding the virus, and optimal policy is likely to be different when there are a lot of unknowns. — PolitiTweet.org

Nate Silver @NateSilver538

@DKThomp My apologies to the people of Denmark for the oversight. — PolitiTweet.org

Nate Silver @NateSilver538

I could go on all day citing research that attempts to quantify these; suffice it to say that the magnitudes are disputed. But just as a stylized example, if you had 5 factors that each reduce transmission by 20%, then a wintertime R0 of 2.5 could equal a current Rt of 0.8. — PolitiTweet.org

Nate Silver @NateSilver538

Take your pick from: - Weather - Masks/hygiene - Reduced susceptibility (i.e. some people had it and are now immune) - Improved testing/tracing/awareness - Still significant (though diminishing) social distancing — PolitiTweet.org

Nate Silver @NateSilver538

Boring take: there could be a lot of little things that combine to reduce spread and which collectively have a fairly large impact, rather than there necessarily being a magic bullet. — PolitiTweet.org

Nate Silver @NateSilver538

@nataliemj10 @gelliottmorris Yeah and also we never have enough data on presidential elections so throwing out most of the data we *do* have to focus only on the 3-5 most recent elections seems like a big mistake. — PolitiTweet.org

Nate Silver @NateSilver538

Some of these "the future of X post-pandemic" pieces are smart, but most would do well to clarify what future they're referring to. Is it one where there's still some heightened risk of COVID-19? Or, is it mostly the psychological and/or economic after-effects of *this* pandemic? — PolitiTweet.org

Nate Silver @NateSilver538

Subjectivity (which you want to avoid) vs. judgment (which all models require) is admittedly a fine line. But generally speaking, you should strongly prefer finding systematic rather than ad-hoc methods for handling "problematic" cases. — PolitiTweet.org

Nate Silver @NateSilver538

Yeah for sure. And trying to manually remove "bad" data points without taking the time to develop a systematic process/philosophy for doing so (i.e. building an algorithm) introduces an awful lot of subjectivity. — PolitiTweet.org

Nate Silver @NateSilver538

@mattyglesias There can also be the issue of overkill, e.g. pollsters introduce 3 or 4 changes to try to combat D bias. It turns out only one of those changes (e.g. education weighting) is correct. The other ones introduce an R bias. That's basically what happened in the UK in 2017. — PolitiTweet.org

Nate Silver @NateSilver538

@mattyglesias But it's not just educational attainment. There are a lot of sources of bias, and if the polling community focuses more on sources that could produce a D bias (which seems *very* likely post 2016), it's not clear where you'd expect the bias of a polling average to end up. — PolitiTweet.org

Nate Silver @NateSilver538

I really appreciate smart folks who are looking at individual polls and pointing out potential flaws. I don't do that so much these days; I tend to stay very focused on the macro. As a macro strategy, "take the weighted average and account for uncertainty" is still hard to beat. — PolitiTweet.org

Nate Silver @NateSilver538

A good example of overcompensation is the 2017 UK election, where after several years of a pro-Labour bias, pollsters started doing some *really* dodgy things to basically just slant their results toward Conservatives and wound up way low on Labour. https://t.co/GTjBmg9OjD — PolitiTweet.org