Nate Silver

Nate Silver @NateSilver538

This includes the possibility of clustered errors, where errors are based on demographic, political or geographical relationships between the states, or errors related to mail voting and COVID-19 prevalence. https://t.co/nVtlpQscbB — PolitiTweet.org

Nate Silver @NateSilver538

There is a long discussion about how we generate these maps in our methodology guide. Our model assumes that the majority of the error (something like 75%, although the amount varies over time) is correlated across states rather than applying to any one state individually. — PolitiTweet.org

Nate Silver @NateSilver538

*Collectively*, the long tail of very weird 1-in-1,000 and 1-in-10,000 scenarios might have a 1-5% chance of occurring, or something on that order, depending on what your threshold is for "very weird", so a few of them will come up when we show a batch of 100 maps. — PolitiTweet.org

Nate Silver @NateSilver538

But it's important to keep in mind that if we happen to show a given map out of the sample of 100 the program randomly selects, that doesn't mean it has a 1 in 100 chance of occurring. The odds of that *exact* scenario might be 1 in 10,000 or 1 in a million, etc. — PolitiTweet.org

Nate Silver @NateSilver538

If you look through these, you can usually find a few maps that look *pretty* weird (say, Biden winning GA but losing FL) and maybe 1 or 2 that look *really* weird. This is how the model is supposed to work accounting for a small chance of fat-tailed errors. — PolitiTweet.org

Nate Silver @NateSilver538

So, every time we run our presidential model, we show 100 maps on our interactive. The maps are randomly selected, more or less, out of the 40,000 simulations we conduct with each model run. People really like looking at these but they can sometimes yield misconceptions. THREAD https://t.co/JceoiXMed7 — PolitiTweet.org

Nate Silver @NateSilver538

I think it probably finds a receptive audience. There are plenty of MAGA bros in the financial sector who often hold somewhat half-baked theories about politics and hold them *very* confidently. Also plenty of Democrats in finance who think "polls always get it wrong". — PolitiTweet.org

Nate Silver @NateSilver538

RT @mattyglesias: Because it’s not unusual for polls to be off by 2-3 points and because Trump plausibly has a 2-3 point edge in the Electo… — PolitiTweet.org

Nate Silver @NateSilver538

@joshtpm Yeah, it's also a somewhat Republican-leaning group of pollsters. YouGov and Morning Consult, which have don't have strong house effects, have the race at 6 and 8 points respectively, so our average showing it at 7 might be about right. — PolitiTweet.org

Nate Silver @NateSilver538

But usually, if someone is obsessed with shy Trump voters, it's a sign that they have a sophomoric understanding of the field. It's the sort of thing that makes you sound smart but doesn't really have much evidence behind it once you've done the research. — PolitiTweet.org

Nate Silver @NateSilver538

Again, one must be aware of the possibility of correlated polling error *in either direction*. I can imagine reasons that polls would underestimate Trump again. Can also imagine some reasons they'd underestimate Biden especially if pollsters are "fighting the last war" from 2016. — PolitiTweet.org

Nate Silver @NateSilver538

Also worth noting that polls in the midterms showed no bias in either direction. Nor did Trump over-perform his polls in the 2016 primaries. Nor do nationalist parties have any history of outperforming their polls in a large sample of European elections. https://t.co/KejtGLldTD — PolitiTweet.org

Nate Silver @NateSilver538

There's not really much/any evidence of shy Trump voters at all. He only beat his polls by ~2 points in 2016, which can pretty easily be explained by not properly weighting by education + late-deciding voters breaking toward him. — PolitiTweet.org

Nate Silver @NateSilver538

While one should always be alert to the possibility of systematic polling errors in either direction, this (alleging that polls are biased by 5-6 points against Trump because of shy Trump voters) is completely insane. https://t.co/ESIk22qjZl https://t.co/AcWwaWojD9 — PolitiTweet.org

Nate Silver @NateSilver538

Overall, though, the set of polls out over the past couple weeks is of somewhat uneven quality, and really no live-caller polls at all. So it's a bit early to come to too many conclusions. https://t.co/9AeTKX5o0O — PolitiTweet.org

Nate Silver @NateSilver538

Morning Consult's national numbers are also interesting in that they initially showed Biden's lead falling to 6 in a one-day poll just after the convention but it has now bounced back to 8. Would possibly suggest that the RNC helped Trump but maybe not the Kenosha news cycle. — PolitiTweet.org

Nate Silver @NateSilver538

Our national polling average is steady at Biden +7.1 after several polls in overnight, while Biden's Electoral College win probably is up just a hair after a good set of state polls for him from Morning Consult: https://t.co/ajG88SznSA — PolitiTweet.org

Nate Silver @NateSilver538

RT @galendruke: 🎧 New pod 🎧 The media is getting ahead of itself on the effects of Kenosha and Portland on the election. https://t.co/Lab… — PolitiTweet.org

Nate Silver @NateSilver538

Perhaps more important than the somewhat esoteric point about models is that the economy could still wind up being an asset for Trump as he may be plausibly able to argue that the country is on its way back, etc. — PolitiTweet.org

Nate Silver @NateSilver538

Seems like you shouldn't have a model that relies too strongly on the assumption that "fundamentals" will doom Trump when third-quarter GDP growth is now forecasted at +29% (annualized) after a -32% in the second quarter. https://t.co/rYehldpZCk. — PolitiTweet.org

Nate Silver @NateSilver538

RT @LPDonovan: https://t.co/z2xoY1dCaf https://t.co/C73I0n5Krm — PolitiTweet.org

Nate Silver @NateSilver538

RT @WesPegden: @NateSilver538 Indeed, because of the fact that child cases (also older-adults) lag the younger-adult demographic which driv… — PolitiTweet.org

Nate Silver @NateSilver538

Separately, these charts would appear to show a spike in cases among children. But what this data actually shows is growth in the *cumulative* number of cases and not the number of *new* cases. That's counterintuitive, maybe even on the verge of misleading. https://t.co/2pRdHsKMjN — PolitiTweet.org

Nate Silver @NateSilver538

A bit weird. I'm not sure why you'd look at the *share* of infected cases among children as opposed to the *number* of cases. If e.g. you close bars and there's less adult transmission, the *share* of cases among children will go up but the *number* won't. https://t.co/fZeFPWsCku https://t.co/e4Uynap49Z — PolitiTweet.org

Nate Silver @NateSilver538

There were many, many arguments about exactly where to draw these thresholds and what words to use. Anything in the ~70% range tends to be pretty tricky to convey, I think. — PolitiTweet.org

Nate Silver @NateSilver538

I think people's intuitions get thrown off by the Electoral College. A 7-point lead in the popular vote (which our model thinks will tighten by another point or so) is decently robust even on Aug. 31. But it's more like a 4-point lead in the tipping point states, which is iffier. — PolitiTweet.org

Nate Silver @NateSilver538

To me, that seems pretty sensible. A 3-point forecasting error for an election that's still 2+ months away is going to be fairly common. There's a contingent on here that's like "why does 538's model show so much uncertainty!" but that doesn't seem like a ton of uncertainty IMO. — PolitiTweet.org

Nate Silver @NateSilver538

This also gives you a bit of insight into how our model works. If, on August 31, a candidate is projected to win in the decisive state by 3 points, how does that translate into a *probability* of victory? He (Biden in this case) wins about 2 out of 3 times, our model is saying. — PolitiTweet.org

Nate Silver @NateSilver538

I guess everybody is just going to go ahead and attribute any polling movement to Kenosha/Portland when the prior should just be that it's a convention bounce, and seemingly not an especially large bounce at that. — PolitiTweet.org

Nate Silver @NateSilver538

If you want to get nit-picky the gap is probably more like 3 points than 4. Our model currently has Biden projected to win the popular vote by 5.7 points (note this assumes some further tightening) while he's projected to win by 2.9 in the average tipping-point state. — PolitiTweet.org