Nate Silver

Nate Silver @NateSilver538

@WinWithJMC @NCSBE Obviously this could have something to do with counties, as the streaks fairly closely align with county boundaries (though not perfectly so). Maybe some counties report their rejected ballots first? Maybe some have issues with ballot design for county-level offices? — PolitiTweet.org

Nate Silver @NateSilver538

@WinWithJMC There are some pretty weird patterns here in terms of which ballots are listed as SPOILED, i.e. instead of being randomly dispersed throughout the list, there are large blocks of several SPOILED ballots in a row. Any ideas about what's going on with this? @NCSBE https://t.co/oMcEWa3ceo — PolitiTweet.org

Nate Silver @NateSilver538

RT @perrybaconjr: More than a third of Trump's supporters in 2016 were people of color (about 10 percent) or non-Hispanic white American… — PolitiTweet.org

Nate Silver @NateSilver538

McCain was probably bound to lose anyway (his lead at this point reflected a Palin/convention bounce, at least in part) but a reminder that basically the entire financial crisis played out between this point in the campaign and the election in 2008. News can still happen. — PolitiTweet.org

Nate Silver @NateSilver538

@TheStalwart FWIW, though, we don't use GDP in our model. 4 of the 6 factors we consider (payrolls, disposable income, inflation, stocks) are stuff that's fairly visible to people. And the 5th (personal consumption expenditures) can reflect consumer confidence fairly well. — PolitiTweet.org

Nate Silver @NateSilver538

@TheStalwart I mean, I think there's always the question of how people are experiencing the economy vs. what the "objective" data says. I think that's an intrinsic issue with economic fundamentals models, and one reason why they aren't all that accurate out-of-sample. — PolitiTweet.org

Nate Silver @NateSilver538

3. Labor Day is an inflection point where polls *begin* to get more reliable. But we're still at just the beginning of the end, if you will. There are still the debates and lots of campaigning and news left to go. 4. Polls are off by an average of ~3 points on Election Day. — PolitiTweet.org

Nate Silver @NateSilver538

1. It's a closer race in the Electoral College, with Biden ahead by perhaps 4-5 points in the tipping-point states. 2. Our model expects the race to tighten by a point or so because of improvement in the economy. (cont.) — PolitiTweet.org

Nate Silver @NateSilver538

Meanwhile, our *forecast* gives Trump a 28% chance of winning, and Biden a 72% chance. So the outcome is a long way from certain, despite Biden's current polling lead. Where do Trump's chances come from? https://t.co/ajG88SznSA https://t.co/o2t9OFGEza — PolitiTweet.org

Nate Silver @NateSilver538

It's after Labor Day! It's no longer "too soon" to look at polls, even you're the type of person who's inclined to be conservative in when you start looking at polls. Joe Biden leads by 7.5 points in our national polling average: https://t.co/cy51vc5isJ https://t.co/1LKeC3qqtW — PolitiTweet.org

Nate Silver @NateSilver538

I understand that there are often good reasons to use anonymous sources, but my guess is that journalists underestimate the degree to which the general public finds "you can't trust anonymous sources!" to be a persuasive argument. — PolitiTweet.org

Nate Silver @NateSilver538

@Noahpinion You can't spell ELECTION FORECAST without N-A-T-E — PolitiTweet.org

Nate Silver @NateSilver538

BTW, there's not just the question of how voters weigh Q3 relative to Q2, but also how they weigh Q1 and 2019, etc. If voters have a lagged time horizon, that is not necessarily bad for Trump. It could also mean they still give him credit for the pre-COVID economy. — PolitiTweet.org

Nate Silver @NateSilver538

But FWIW, based on settings we chose 8 year ago that weight more recent quarters *slightly* more heavily, our model sees the economy as a roughly neutral factor for Trump. We also do not weight the economy very much, though, precisely because of specification issues like these. — PolitiTweet.org

Nate Silver @NateSilver538

To answer @paulkrugman's question: yeah, there's nowhere near enough data to know what a record decline in Q2 followed by a record rebound in Q3 is supposed to do to an incumbent's re-election chances. So be careful about confident proclamations about this. — PolitiTweet.org

Nate Silver @NateSilver538

RT @maggiekb1: Adding that, from what experts tell me, no model should be used to make projections beyond 3-4 weeks out. — PolitiTweet.org

Nate Silver @NateSilver538

I think it's good to say there's *potential* for a fall spike in COVID cases. But something gets lost in translation and the headline and lede make it seem like there's a consensus of experts definitively predicting a fall wave, when I don't believe there is any such consensus. — PolitiTweet.org

Nate Silver @NateSilver538

Can't say I'm in love with articles like these. Models are all over the place for what to expect from COVID in the fall, and have wide uncertainty ranges. The article heavily cites the IHME model while ignoring others that have been more accurate so far. https://t.co/Knw5NwoWkN — PolitiTweet.org

Nate Silver @NateSilver538

The other issue with an obsessive focus on WI and MN is that they simply don't have very many electoral votes, only 10 each. — PolitiTweet.org

Nate Silver @NateSilver538

Well, then he's liable to gain the most ground relative to Clinton in states that had lots of Obama-Trump voters, such as Wisconsin, Iowa and Maine. I'd say the polling mostly bears this out, though some of the states we'd want to look at haven't been polled much lately. https://t.co/Fjb2SPEgwt — PolitiTweet.org

Nate Silver @NateSilver538

In some ways, the answer to "why is Biden polling so well in Wisconsin?" may be surprisingly simple. Let's say Biden is able to win back, I don't know... 40% of Obama-Trump voters. FWIW, that's about what it would take to get him to +7/8 nationally, assuming nothing else changes. — PolitiTweet.org

Nate Silver @NateSilver538

RT @DavidLJarman: I'm worried you may not be talking to enough people in enough different locations? It's too bad there's no way to use a c… — PolitiTweet.org

Nate Silver @NateSilver538

@Nate_Cohn It's very close to 50/50 though (Trump's small edge comes from incumbency at this point.) Err I should say in the popular vote. I can see an argument that the prior should be calibrated off the tipping-point state instead. — PolitiTweet.org

Nate Silver @NateSilver538

@Nate_Cohn Yeah that's basically what the prior is trying to say, in our view (i.e. how would a generic elected incumbent be doing under these conditions). There's a sense in which it feels weird to treat Trump as a generic *anything*, but I'm not sure how to operationalize that in a model. — PolitiTweet.org

Nate Silver @NateSilver538

@Nate_Cohn I feel pretty good about our prior (which has drifted up to around Trump +2 nationally after further good economic data) FWIW. His approval on the economy is good. Q3 GDP numbers could be nearly as good as Q2 was bad. Incumbents around the world have mostly gotten more popular. — PolitiTweet.org

Nate Silver @NateSilver538

For all the focus on Wisconsin, Biden has been polling conspicuously well there as compared to the other swing states. Currently +7.1 on quite a bit of recent polling, which is pretty similar to his national lead (+7.5). https://t.co/C8GYebMykJ https://t.co/zf3Tp7eefc — PolitiTweet.org

Nate Silver @NateSilver538

I don't think there's always a huge amount of info in how campaigns themselves assess the state of the race. But still, it's pretty remarkable if it's only Labor Day, and allies of Trump think there's "virtually no chance" he wins the popular vote. https://t.co/40JYaBl1Yk https://t.co/ZOHdXOzrfF — PolitiTweet.org

Nate Silver @NateSilver538

@skepticalsports Yeah, I think these markets are dumber the bigger the event gets. — PolitiTweet.org

Nate Silver @NateSilver538

I don't know how the contracts at these markets are written to account for cases where, say, Trump doesn't concede the election. Apart from something like that, it doesn't make any sense to have Biden in the 60s in the tipping-point states but only 53% to win nationally. — PolitiTweet.org

Nate Silver @NateSilver538

Another thing that doesn't make any sense about betting markets is that their state-by-state odds are very close to 538's projections but their national projections are quite different. https://t.co/qpmkmQfImF https://t.co/Yy9OXpC5N5 — PolitiTweet.org