Nate Silver

Nate Silver @NateSilver538

@databyler 🎵 Herding time, you don't have to poll But you can't publish a D+7 generic ballot 🎵 Sorry, I think I need to workshop that one a bit more. — PolitiTweet.org

Nate Silver @NateSilver538

(More on herding here: https://t.co/mIFXJyFxXs) — PolitiTweet.org

Nate Silver @NateSilver538

Spent 15 minutes trying to compose a song called "herding time" to the tune of "jingle bells" but decided that wasn't a good use of my time. But man, pretty much every poll today is sticking very close to the averages. Kudos to those who got a weird result and ran with it. — PolitiTweet.org

Nate Silver @NateSilver538

@karlbykarlsmith This one's more unusual in that it doesn't reflect any sort of intentional herding, I don't think. As far as I'm aware, these are all “set it and forget it" forecasts where people aren't making any manual tweaks. — PolitiTweet.org

Nate Silver @NateSilver538

@mattyglesias 5.38 million times — PolitiTweet.org

Nate Silver @NateSilver538

Exciting to see the wild divergence in the algorithmic Senate forecasts this year. Do Republicans have a 54% chance, a 55% chance, or a 56% chance? #ModelWars https://t.co/9C7sfLzB4x — PolitiTweet.org

Nate Silver @NateSilver538

RT @NateSilver538: Our biennial PSA to pollsters: Barring exceptional circumstances, we'll run our final 538 midterm forecast just after m… — PolitiTweet.org

Nate Silver @NateSilver538

Here are the midterm questions that keep me up at night (warning: fairly wonky/nerdy post). https://t.co/GUpmaUjKnK — PolitiTweet.org

Nate Silver @NateSilver538

@RossBarkan She's about as likely to outperform her polling as to underperform it. Those big polling errors in recent years haven't tended to be in blue states. — PolitiTweet.org

Nate Silver @NateSilver538

@billscher @lxeagle17 Polls do get entirely phased out in our average at some point. How quickly to phase them out is an empirical question and is different depending on the type of poll. Generic ballot polling movement tends to be noisy with large house effects so it's best to phase out slowly. — PolitiTweet.org

Nate Silver @NateSilver538

Generally likely voter polls haven't been great for Democrats, but not much of a likely vs. registered voter gap in the final ABC/WaPo and NBC polls. Quite possibly the difference between them keeping or losing the Senate. https://t.co/cS3SAvi2sQ — PolitiTweet.org

Nate Silver @NateSilver538

@willwilkinson Modeling and forecasting is hard (and interesting) precisely because it has to wrestle with this challenge. — PolitiTweet.org

Nate Silver @NateSilver538

@willwilkinson Kind of a lazy argument TBH! — PolitiTweet.org

Nate Silver @NateSilver538

New rule: you're not allowed to criticize any pollster or polling aggregator unless you pre-register your own forecast. Use this thread to predict the following: 1—% chance of R Senate win 2—% of R House win 3—# of R Senate seats 4—# of R House "" 5—GA-Sen result 6—AZ-Gov result — PolitiTweet.org

Nate Silver @NateSilver538

Life is a lot easier even you can just stick to vague pronoucements and innuendo instead of devoting 1000s of hours to developing a rigorous process that people can test against actual results. — PolitiTweet.org

Nate Silver @NateSilver538

Well, that's one of the whole purposes of 538. Our pollster ratings track accuracy and improve pollster incentives to put out accurate numbers rather than just influencing the discourse. If Trafalgar et al have a bad night, that will lower their rating going forward. — PolitiTweet.org

Nate Silver @NateSilver538

Part of the reason for building a model and polling averages with **hard and fast, deliberately inflexible rules** is because I don't want to have to make decisions in the heat of the moment when the chance of motivated reasoning is high. — PolitiTweet.org

Nate Silver @NateSilver538

I get that we all engage in motivated reasoning, trust me I do it too, but the propensity for motivated reasoning when looking at polling and other data is **very** high the week or so before an election. — PolitiTweet.org

Nate Silver @NateSilver538

(As an aside, whenever someone uses weasel words like "basically admits" they're almost always mischaracterizing someone's actual argument.) — PolitiTweet.org

Nate Silver @NateSilver538

That's NOT what I'm saying at all. There are strong incentives for pollsters to be accurate. If people think the polling out there are inaccurate, they should publish more accurate polls. — PolitiTweet.org

Nate Silver @NateSilver538

@yeselson @LPDonovan @ThePlumLineGS Yeah, Trafalgar's not always been super transparent about which polls they're doing on their own, and which are for R clients, so they're something of a special case. https://t.co/i9M7cKLc6x — PolitiTweet.org

Nate Silver @NateSilver538

@Chris_arnade Yeah and in some ways, the more of a political insider you are, the less you tend to think like a rank-and-file voter. — PolitiTweet.org

Nate Silver @NateSilver538

@yeselson @LPDonovan @ThePlumLineGS Empirically the most accurate strategy is giving polls a bonus (a higher weight in our average) if they join a transparency initiative but not throwing them out if they don't, so that's what we do. Partisan polls also get a lower weight and a bigger house effect adjustment. — PolitiTweet.org

Nate Silver @NateSilver538

Political insiders tend not to have well-trained intuitions for forecasting, in part because elections happen infrequently. And they tend to be quite conformist relative to their social circles. They are usually regurgitating vibes more than revealing any proprietary knowledge. — PolitiTweet.org

Nate Silver @NateSilver538

Yeah, people forget that insider buzz about the race is not only usually pretty worthless but in the past has often been a contrarian indicator, e.g. you would have made money by betting on the opposite of what the insiders think. https://t.co/tyIZ8crzGP — PolitiTweet.org

Nate Silver @NateSilver538

@AGHamilton29 Right yeah and in fact Trump would frequently allege that "fake polls" were part of the conspiracy. https://t.co/jDD38RzpzY — PolitiTweet.org

Nate Silver @NateSilver538

@AGHamilton29 The other version (which I am very much not endorsing, to be clear) is more openly conspiratorial: that these polls are designed to provide cover for candidates to dispute election results later on ("all the polls had us ahead"). — PolitiTweet.org

Nate Silver @NateSilver538

@AGHamilton29 There are two versions of this coming from Ds. One version is that it designed to lower D morale. It's not clear this makes much sense since you could just as easily argue it the other way (e.g. tighter polls reduce D complacency). — PolitiTweet.org

Nate Silver @NateSilver538

@Protentialmn Well, yeah, lots of key races within just a couple of percentage points. Very little is locked in yet! — PolitiTweet.org

Nate Silver @NateSilver538

The barriers to entry right now in polling are fairly low. If someone thinks the averages are wrong because other firms are "flooding the zone", they can do their own polling and will get to look smart and get lots of business in 2024/26 if they're right. — PolitiTweet.org