Nate Silver

Nate Silver @NateSilver538

@studentactivism @jonathanchait @owasow I'm not the expert here but it seems like when the protests are sort of inchoate, that's when they can turn more violent or at least have the violent elements highlighted in news coverage, but when there's critical mass around them, they tend to be more peaceful and popular. — PolitiTweet.org

Nate Silver @NateSilver538

RT @mattyglesias: Just in case anyone needs some better grounding for their “don’t count Trump out!” columns, I wrote something up. https:/… — PolitiTweet.org

Nate Silver @NateSilver538

I don't mean to be too flip here. I think it's appropriate to say there are a lot of uncertain factors in the race, including the protests. But many predictions about the impact of the protests were wrong back in May/June. Better to wait for public opinion data this time around. — PolitiTweet.org

Nate Silver @NateSilver538

Here is a result that "Kenosha will help Trump win!" pundits may be surprised by. The NBA boycott/protest/strike around Jacob Blake was very popular. — PolitiTweet.org

Nate Silver @NateSilver538

RT @baseballot: For an idea of how distorted the returns on election night could be, consider that Trump might win the in-person Election D… — PolitiTweet.org

Nate Silver @NateSilver538

Yeah, I think this is part of it. Reporters are hearing Trump's argument, and the conventions are the one time when there isn't the immediate opportunity to put a fact-check or context around it. But most of it won't be particularly new to actual voters. https://t.co/Yhn0rrn6xx — PolitiTweet.org

Nate Silver @NateSilver538

There are also constant admonitions not to count Trump out... but almost nobody is counting Trump out. I do think some of the non-538 models are overconfident on Biden, which is important in MY world but a pretty narrow category of election coverage overall. — PolitiTweet.org

Nate Silver @NateSilver538

Overall, we're at a weird moment in the campaign and the conventional wisdom surrounding it. There is a perception that things are shifting toward Trump far more than there is much in the way of evidence to support that conclusion. — PolitiTweet.org

Nate Silver @NateSilver538

What we still do not have, unless I missed something, is *any* live-caller telephone national polls since the Democratic convention. Live-caller polls tend to show bigger swings ... so it's hard to come to too many conclusions about bounces from online polls alone. — PolitiTweet.org

Nate Silver @NateSilver538

USC Dornsife (affiliated with the LA Times in 2016) is different than most other polls in that they interview the same panel of people continually. They also have 2 versions of their topline (one probabilistic and one deterministic); we use a mix of both versions in our averages. — PolitiTweet.org

Nate Silver @NateSilver538

So now we've got a couple of polls out that *do* appear to show a bit of a convention bounce for Biden. Navigator shows him at +13 post-conventions vs. +10 earlier in the month. And the USC Dornsife tracking poll is back and shows a ~3 point bounce. https://t.co/cy51vc5isJ — PolitiTweet.org

Nate Silver @NateSilver538

RT @geoffreyvs: Amazing how so many people seem to know how the GOP convention or the Kenosha situation will affect the polls. I would advi… — PolitiTweet.org

Nate Silver @NateSilver538

The problem with the RNC sending a message that COVID is over is not just that COVID isn't over (1,100 deaths were reported today) but also that there's very much still time for another resurgence by Nov. 3, and such a resurgence is more likely if people think it's over. — PolitiTweet.org

Nate Silver @NateSilver538

If he loses and/or doesn't get a big convention bounce, won't it be seen as having been ineffective and deter other presidents from using it in the future? — PolitiTweet.org

Nate Silver @NateSilver538

Not sure if this is good news or bad news. But this is our *last* day of liveblogging the conventions. It's Day 4 of the RNC. https://t.co/rxKH6XJY6s — PolitiTweet.org

Nate Silver @NateSilver538

¯\_(ツ)_/¯ https://t.co/WSUgXAyafY — PolitiTweet.org

Nate Silver @NateSilver538

RT @aedwardslevy: @NateSilver538 in fairness it is probably a more representative metric than diner interviews — PolitiTweet.org

Nate Silver @NateSilver538

I assume he's correct that Facebook has become a go-to news source for a lot of conservatives. That's exactly why you CAN'T use it as a substitute for polls! It's not representative! It's *exactly* like saying "I watch lots of Fox News, so I know Trump's going to win." — PolitiTweet.org

Nate Silver @NateSilver538

You'd think this sort of analysis (which is based on approximately zero evidence) would have become less common after the very NOT online Joe Biden crushed all comers in the Democratic primary. https://t.co/eNCzS0iX1Q https://t.co/CZnx4MKgqh — PolitiTweet.org

Nate Silver @NateSilver538

@conorsen Of course it's possible! But as a matter of practice, it's way, way easier to build overconfident models than under-confident ones, especially given the nature of the data in a presidential election forecast (low n, correlated errors). — PolitiTweet.org

Nate Silver @NateSilver538

So when I see data-oriented folks get overly obsessed with what this or that indicator would have said in 2016 or otherwise fighting the last year, I think they're starting to commit some of the same sins as conventional punditry with its strong recency bias. — PolitiTweet.org

Nate Silver @NateSilver538

IMO one of the biggest advantages of models is looking at the broader history of elections and not just the most recent or salient examples. There were decent precedents for how Biden won the D nomination (say, Clinton '92) for instance, but they weren't especially recent. — PolitiTweet.org

Nate Silver @NateSilver538

Yeah, the Clinton stuff is sort of like superstition masquerading as empiricism. There is, obviously, nothing wrong with looking at past races to figure out how much polls might change. That's exactly what models do! But at least they're using 12-15 elections and not just 1. — PolitiTweet.org

Nate Silver @NateSilver538

p.s. The biggest tip-off that someone *thinks* they're following the science when they're *not* actually following the science is when they say "immunity only lasts 3 months!" which is a pretty incorrect & outdated caricature of what the best research now says. — PolitiTweet.org

Nate Silver @NateSilver538

More truth in this than a lot of people will want to admit! — PolitiTweet.org

Nate Silver @NateSilver538

RT @dylanlscott: A pretty convincing case from @perrybaconjr that public attitudes toward Black Lives Matter and the electoral prospects of… — PolitiTweet.org

Nate Silver @NateSilver538

@jbarro There's a discussion in here. Convention bounces have declined in general, likely because of higher partisanship, but it could also reflect other factors (e.g. changes in polling practices). The model assumed there would be rather small bounces this year. https://t.co/nVtlpQscbB https://t.co/qCzgjUH4tp — PolitiTweet.org

Nate Silver @NateSilver538

I don't know if there's a *consensus*, but my impression of the research on this topic is that the single *most* important factor is just that SARS-CoV-2 was introduced very early in NYC (probably early February), perhaps with multiple introductions. https://t.co/FJSAJeXRDU — PolitiTweet.org

Nate Silver @NateSilver538

That's not to say there are no risks here for Democrats. Maybe there is some degree of quiet backlash building to the protests (although note that support for BLM is still equal/higher than pre-protests). But if Trump tries to loudly exploit that, it could backfire. — PolitiTweet.org

Nate Silver @NateSilver538

This is relevant too. In Wisconsin, while there's been a big decline in support for BLM, evaluations of Trump's handling of the protests is just as bad as before. https://t.co/kb1tMrizOl — PolitiTweet.org