Nate Silver

Nate Silver @NateSilver538

If you look closely, both Biden and Trump are gaining ground at the expense of undecided; 51.3% is Biden's highest number of the year, and 42.9% is Trump's high since the George Floyd protests, though he was as high as 45.6% pre-protests and pre-COVID. https://t.co/aWp4iSMyh9 — PolitiTweet.org

Nate Silver @NateSilver538

Is the race tightening? I don't know. Biden peaked in our average at 9.6 points at various times in late June and early July. Our averages are pretty smart, but a 1.2-point shift is sort of on the fringe between what you'd consider real movement versus noise. — PolitiTweet.org

Nate Silver @NateSilver538

...and back up to +8.4 with the addition of the ABC/Post poll. https://t.co/cy51vc5isJ https://t.co/U2KQmYLXhS — PolitiTweet.org

Nate Silver @NateSilver538

@dandrezner Some pollsters don't even bother to call boatphones. — PolitiTweet.org

Nate Silver @NateSilver538

Was randomly in New Canaan, CT today (very upscale NYC suburb with lots of hedge fund wealth, etc.) and ... did any other town of comparable size shift this much toward Democrats from 2012 to 2016? Went from Romney +29 to Clinton +11. https://t.co/gS2A56n18f — PolitiTweet.org

Nate Silver @NateSilver538

THE MEDIA OVERREACTS TO OUTLIER POLLS™ critique is sometimes true, but the press has gotten more sophisticated in looking at polling averages, etc. So while overreactions are not uncommon, there are also plenty of proper reactions, and even the occasional under-reaction. — PolitiTweet.org

Nate Silver @NateSilver538

...and back down to +8.0 after adding the new CNN poll. https://t.co/cy51vc5isJ https://t.co/dGwrqBcgei — PolitiTweet.org

Nate Silver @NateSilver538

Biden's lead is now 8.6 points in our national polling average. Similar to or very slightly higher than the past few weeks, when it was hovering around ~8.0. At least one more big poll (@ABC / WaPo) coming out before the conventions start tomorrow. https://t.co/cy51vc5isJ https://t.co/dcofxsDEnK — PolitiTweet.org

Nate Silver @NateSilver538

@WesPegden Yeah, this dynamic comes up with a lot of pandemic-related stuff! Also maybe with respect to stuff like which businesses fail. — PolitiTweet.org

Nate Silver @NateSilver538

As a corollary, if you're getting rid of people who would have flaked on NYC anyway, the people who remain are likely to be a pretty hardy lot, which could help with the city's resiliency in 2021 and beyond. — PolitiTweet.org

Nate Silver @NateSilver538

It seems probable that many of the people who will permanently leave NYC during the pandemic are people who were likely to leave within 3-5 years anyway. — PolitiTweet.org

Nate Silver @NateSilver538

I am also not sure* that it is necessarily good to have such a distinction ^^ but I think public health is uniquely complicated in some ways, especially in the midst of a pandemic, and generally operates on the presumption that there is one. * typo fixed — PolitiTweet.org

Nate Silver @NateSilver538

I am also sure that it is necessarily good to have such a distinction ^^ but I think public health is uniquely complicated in some ways, especially in the midst of a pandemic, and generally operates on the presumption that there is one. — PolitiTweet.org

Nate Silver @NateSilver538

I think part of the problem is that the "scientists talking to the public" and "scientists talking to other scientists" channels are harder to keep separate than they once were, and that the public sometimes intercepts messages intended for the scientists-to-scientists channel. — PolitiTweet.org

Nate Silver @NateSilver538

...the most obvious one being reducing the partisan split in intention to cast votes by mail. But it could also lead to Democratic voters being more diligent overall, e.g. requesting their ballots earlier and sending them in sooner, exploring in-person early voting, etc. — PolitiTweet.org

Nate Silver @NateSilver538

One other thing that's complicated here is that if Democrats are reading mainstream media sources that are covering this story, while Republicans are reading "alternative"/right-wing sources that largely aren't, that could have various effects... — PolitiTweet.org

Nate Silver @NateSilver538

This is the sort of story that shows the importance of a free press. (At the same time, probably time to end the "Why is nobody talking about the Postal Service?!?" meme. It's—rightly— become a huge story.) — PolitiTweet.org

Nate Silver @NateSilver538

@mattyglesias Yeah, I don't know, and there's at least some evidence to suggest the opposite, i.e. that fatalism makes people want to give up on control measures (I don't think the paper below proves its theory, but it's at least plausible.) https://t.co/JY5O4698zD — PolitiTweet.org

Nate Silver @NateSilver538

The other frequent culprit behind "unexpected" swings is a big change in one of the variables we use in our economic index, including the stock market. We'll soon have ways for you to download the economic index so you can see when it has changed. OK, boring thread over. — PolitiTweet.org

Nate Silver @NateSilver538

Plus keep in mind that outliers do happen. One in 20 cases fall outside the margin of error. We update the model several times each day; you're going to have a couple of outliers a week. Bottom line: do not sweat a change unless it's AT LEAST 2 full points, e.g. from 30% to 32%. — PolitiTweet.org

Nate Silver @NateSilver538

This is one reason why we aren't listing the decimal, i.e. why we list Trump's win probability as 28% rather than 28.2%. Of course, this leads to another issue. The probability could change from say 27.51% to 27.49% and it would like it had shifted a full point (from 28% to 27%). — PolitiTweet.org

Nate Silver @NateSilver538

Boring post. But if you're asking "why did 538's numbers change?", keep in mind that there is a tiny bit of sampling error. We run 40,000 simulations with each model run, which translates to a margin of error of around* 0.5%. * Less than that for probabilities close to 0 or 100 — PolitiTweet.org

Nate Silver @NateSilver538

Half of the miscommunications have been caused by misunderstanding how the public would understand certain jargon or other terms of art. The other ½ have been caused by half-baked theories about how the public might react, e.g. concerns about a false sense of security or alarm. — PolitiTweet.org

Nate Silver @NateSilver538

@joshtpm @CT_Bergstrom There is also something about the sense of "not wanting to give people a false sense of security" that has led to incredibly awful and confusing coverage on this issue. — PolitiTweet.org

Nate Silver @NateSilver538

RT @CT_Bergstrom: This bullshit about immunity lasting only three months is everywhere tonight. It's a total misunderstanding. What the CD… — PolitiTweet.org

Nate Silver @NateSilver538

RT @baseballot: FiveThirtyEight is out with a big project this week—no, not that one. Introducing our state-by-state guide to voting in the… — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris @DanRosenheck @SethS_D I mean my personal view is that the disruption in UI could be a big deal, but the SP500 and the economic forecasting panels don't seem to agree, and that's what our model uses. — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris @DanRosenheck @SethS_D We don't actually use GDP at all, in part because it updates quarterly and we want monthly metrics. Of the 6 components we use, one is already very GOOD for Trump (income, because of the CARES Act), two (stocks, inflation) are pretty good, and the other 3 are AWFUL but improving. — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris @DanRosenheck @SethS_D I don't see how either of those things would do an adequate job of accounting for an economy that is projected to get sharply better, in a way that doesn't really resemble a normal recession. — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris @DanRosenheck @SethS_D Especially if you're going to heavily weight the prior, we think it is a pretty big mistake not account for expected changes in the economy given what 3Q GDP growth is expected to be. https://t.co/4a38UV7IQJ — PolitiTweet.org