Deleted tweet detection is currently running at reduced
capacity due to changes to the Twitter API. Some tweets that have been
deleted by the tweet author may not be labeled as deleted in the PolitiTweet
interface.
Showing page 268 of 910.
Nate Silver @NateSilver538
This is far. Though, there's maybe a sense in which voters were taking Trump "seriously but not literally" (sorry) in 2016 and might take the claims more literally now that he's president and has made them so persistently. — PolitiTweet.org
Benjy Sarlin @BenjySarlin
This has some logic to it, but it was also a common take in 2016 when he did the exact same thing the whole month b… https://t.co/SR7dIVAcZ7
Nate Silver @NateSilver538
If Trump keeps implying (or just stating outright) that the election will be rigged, one wonders if some of his supporters might not bother to show up. — PolitiTweet.org
Nate Silver @NateSilver538
RT @galendruke: NEWS: GOP Rep. Will Hurd told me today that he now doesn't know whether he'll vote for Trump this fall. When asked about d… — PolitiTweet.org
Nate Silver @NateSilver538
It's not the most important check on power by any means. But representative and accurate polling does have an important role if there are leaders who would misleadingly declare themselves winners before the vote is counted, or who would try to prevent the vote from being counted. — PolitiTweet.org
Nate Silver @NateSilver538
@jonmladd I'm arguing that models like e.g. the Economist's that have Biden at 99% to win the popular vote based largely on an economic prior are likely misspecified. — PolitiTweet.org
Nate Silver @NateSilver538
I'm not opposed to re-examining assumptions when faced with an obvious edge case. In the abstract, it would be better to have a model that anticipated edge cases in advance—but that's not always feasible. I am, however, opposed to overconfident models ESPECIALLY in an edge case. — PolitiTweet.org
Nate Silver @NateSilver538
If it's a very out-of-sample data point (which it is, although less so if you also use pre-WW2 data) then **nobody has any idea** how the economy impacts the election this year. So either your prior needs to be very weak, or maybe you just need to use a "polls-only" model. — PolitiTweet.org
Jonathan Ladd @jonmladd
Seems like we just had a linear model where a linear approximation worked. But now we have something very out of sa… https://t.co/3x2ZwJ9lzv
Nate Silver @NateSilver538
@gelliottmorris It's not sniping. These are sharp, substantive critiques from someone who has spent more than 12 years now thinking deeply about this stuff. These models are in the public domain and my followers learn something when I make these critiques. So I'm not going to "take it offline". — PolitiTweet.org
Nate Silver @NateSilver538
@gelliottmorris So that's your personal belief? That starts to get a lot different from 100:1. — PolitiTweet.org
Nate Silver @NateSilver538
@gelliottmorris On the off-chance our respective employers would allow it, which they almost certainly wouldn't in my case, could I get some Trump wins the popular vote action from you at 100:1? Or even say 40:1? — PolitiTweet.org
Nate Silver @NateSilver538
Maybe these adjustments are reasonable. But what you *can't* do is say none of the data is really salient to 2020, so much so that you have to make some ad-hoc, one-off adjustments, and *then* claim Biden is 99% to win the popular vote because the historical data proves it. — PolitiTweet.org
Nate Silver @NateSilver538
Literally that's what people (e.g. The Economist's model) is doing. They're just "adjust[ing]" their "economic index" in arbitrary ways so that it doesn't produce a crazy-looking number for 2020. https://t.co/d4B9XWmmw1 https://t.co/gDf3rE4Bkd — PolitiTweet.org
Nate Silver @NateSilver538
Of course, if you're constructing such a model *this year*, you'll choose variables in way that just so happens to come up with a reasonable-looking prediction (i.e. it has Trump losing, but not by 36 points). But that's no longer an economic model; it's just your personal prior. — PolitiTweet.org
Nate Silver @NateSilver538
@mattyglesias People will totally do this. They'll add a "pandemic dummy variable" to 2020 which is literally just mathematically the same as throwing 2020 out of the sample. — PolitiTweet.org
Nate Silver @NateSilver538
Looking at these "fundamentals" models (something I spent a ton of time in 2012) was really a seminal moment for me in learning about p-hacking and the replication crisis. It was amazing to me how poorly they performed on actual, not-known-in-advance data. https://t.co/uzhRzj9s2F — PolitiTweet.org
Nate Silver @NateSilver538
It certainly helps to use a wider array of indicators over a longer time frame (that's what we do) but the notion that you can make a highly *precise* forecast from fundamentals alone in *this* economy just doesn't pass the smell test. — PolitiTweet.org
Nate Silver @NateSilver538
On the other hand, if you ground your model in Real Disposable Income like some others do (e.g. the "Bread and Peace" model) you may have Trump winning in an epic landslide. Possibly the same if you use 3Q GDP (forecasted to be +15% annualized) instead of 2Q. https://t.co/ZX46iLJWCc — PolitiTweet.org
Nate Silver @NateSilver538
Offered as an illustration of the limits of "fundamentals"-based election models: The "Time for Change model", as specified here, predicts that Trump will lose the popular vote by ~36 points (not a typo) based on that 2Q GDP print. https://t.co/2dCpPFMItU — PolitiTweet.org
Nate Silver @NateSilver538
@ForecasterEnten 9/11. The Space Shuttle Challenger. The Berlin Wall falling. The Tim Kaine VP pick. One of those events where you'll always remember where you were. — PolitiTweet.org
Nate Silver @NateSilver538
Cases have begun to plateau—and the positive test rate has gone slightly downward—so you might expect deaths to follow eventually, if the progress is maintained. But for now, we're still seeing week-over-week increases in deaths nearly every day. — PolitiTweet.org
Nate Silver @NateSilver538
Excluding a day with a reporting anomaly, that is the most deaths reported in a day since May 15. Though Texas's numbers may have been influenced by some backlog, quite a few states reported high numbers so it would have been a bad day either way. https://t.co/mAGxHqcmlD — PolitiTweet.org
The COVID Tracking Project @COVID19Tracking
Texas has changed their method of counting deaths, and Hurricane Hanna has hit the state. This may have caused some… https://t.co/LWY5uGD9IN
Nate Silver @NateSilver538
US daily numbers via @COVID19Tracking: Newly reported deaths Today: 1,418 Yesterday: 1,121 One week ago (7/22): 1,117 Newly reported cases T: 66K Y: 54K 7/22: 69K Newly reported tests T: 840K Y: 733K 7/22: 796K Positive test rate T: 7.9% Y: 7.3% 7/22: 8.7% — PolitiTweet.org
Nate Silver @NateSilver538
Namely, it seems like there are a lot of people who are not very politically engaged, but who are also pretty fed up with things. A lot of them probably won't vote, and so they tend to drop off if you use a likely voter screen. But if they do vote, it probably won't be for Trump. — PolitiTweet.org
Nate Silver @NateSilver538
Not a huge sample, but interesting that undecideds go 2:1 for Biden in this poll when required to make a choice, and I think this may have something to do with why we've seen large likely voter vs. registered voter splits in some recent polls. https://t.co/fDUK0aWIiw https://t.co/L0gdSg1B3t — PolitiTweet.org
Nate Silver @NateSilver538
Concerns about "shy Trump voters"—which don't have a lot of support in the evidence—stem from the same place as the impulse to send reporters to diners in Ohio to interview Trump voters, i.e. anxiety among coastal elites that there's a hidden America they can't see or understand. — PolitiTweet.org
Nate Silver @NateSilver538
RT @ryanabest: Sure, Blake Griffin might have been a "stinker" this year — but he's MY stinker and I love him. I scraped @bball_ref to see… — PolitiTweet.org
Nate Silver @NateSilver538
@PollsterPatrick Well, we've tested it empirically! And it's much, much better to trend-adjust state polls than not to do so for the presidential race. You have to be a lot more careful with races for Congress, where the generic ballot doesn't always translate 1:1 into individual races. — PolitiTweet.org
Nate Silver @NateSilver538
In theory, you can also impute these trends from state polls (i.e. if Biden is gaining in MI, he's probably gaining in WI) and our model does do that too. But national polls give you a high volume of polls from a relatively constant set of pollsters—making trend detection easier. — PolitiTweet.org
Nate Silver @NateSilver538
In other words, I don't really care if Biden us up 8 nationally or whatever. I care about the numbers in Wisconsin, Florida, etc. But if Biden has gained 2 points in national polls, it's likely that he's also gained ~2 points in Wisconsin & Florida since they were last polled. — PolitiTweet.org
Nate Silver @NateSilver538
Here's how our model uses national polls. And perhaps it's a good heuristic for other people might think about national polls, also. Pay attention to the trend, not to the level. — PolitiTweet.org
