Nate Silver

Nate Silver @NateSilver538

RT @BenjySarlin: Democrats are banking on their advantage with elderly white voters to rescue them from Trump’s appeal to black and Hispani… — PolitiTweet.org

Nate Silver @NateSilver538

RT @FiveThirtyEight: Here is a list of where Republicans are most likely to pick up Senate seats: https://t.co/xj2w9L8bDR https://t.co/yF… — PolitiTweet.org

Nate Silver @NateSilver538

RT @FiveThirtyEight: Here is a list of Republican-held seats where Dems have at least a 5% chance of winning in at least one version of our… — PolitiTweet.org

Nate Silver @NateSilver538

If you're hungry for even more Senate Model content, we've also got a Model Talk up for ya: https://t.co/S0dRRjgtLs — PolitiTweet.org

Nate Silver @NateSilver538

Why uphill? Democrats *probably* need 4 or 5 pickups, depending on the presidency, since Doug Jones is an underdog in AL-Sen. A lot of these races are against incumbents in red states. Still, Democrats have so many opportunities (~12!) that they're slight favorites to do it. — PolitiTweet.org

Nate Silver @NateSilver538

As charmingly explained here, the difference in the model versions reflects the fact that while Democrats' polling in individual Senate races is quite strong, fundamentals and expert ratings reflect a somewhat uphill battle. https://t.co/ByySuGC8cD — PolitiTweet.org

Nate Silver @NateSilver538

Democrats are slight favorites (58%) to win the Senate according to the "Deluxe" version of our model, which we are treating as the default this year. Their advantage is a bit heavier in the "Classic" (64%) and "Lite" (68%) versions. https://t.co/eev4w6wba9 https://t.co/1aL1HnQXdk — PolitiTweet.org

Nate Silver @NateSilver538

Our SENATE MODEL is LIVE, folks. https://t.co/eev4w6wba9 — PolitiTweet.org

Nate Silver @NateSilver538

Probably a bad sign when an there's an account that feels like the output of "I forced a bot to watch 1000 hours of election modeling debates". — PolitiTweet.org

Nate Silver @NateSilver538

@ForecasterEnten Did Tails wear hipster glasses and memorize obscure facts about elections? Don't think so. — PolitiTweet.org

Nate Silver @NateSilver538

No spoilers but Fivey kind of trolling Martha McSally right here. https://t.co/VGuNCLxqzi — PolitiTweet.org

Nate Silver @NateSilver538

RT @galendruke: The “Thursday” podcast is coming out tomorrow, because... That is when our Senate forecast launches!!! — PolitiTweet.org

Nate Silver @NateSilver538

The meta-debates about election forecasting are so tiresome so let me instead say something randomly positive about @Nate_Cohn / the Upshot which is I really like how they've carved out niches (e.g. the Needle, their own polling) that other data types weren't pursuing so much. — PolitiTweet.org

Nate Silver @NateSilver538

The fact that re-opening, especially of schools, became so polarized along political lines is likely contributing to stupid policy decisions on all sides of the equation. — PolitiTweet.org

Nate Silver @NateSilver538

You could get some mileage, I guess, by comparing the poll says to what it said ~14 days ago. But that still makes it a lagging indicator. In 2016, they polled people once every 7 days so the entire panel was represented in their 7-day window, a far more logical/coherent design. — PolitiTweet.org

Nate Silver @NateSilver538

So effectively, it's as though they have *two* panels rather than one, that get cycled in and out for one another over a 14-day period (plus some additional lag). Perhaps predictably, the poll is following an oscillating pattern so far over a ~14-day cycle. https://t.co/l53cqaMZe1 — PolitiTweet.org

Nate Silver @NateSilver538

But they're implementing it in a very strange way. While their poll shows people who responded over the past *7* days, people are only asked to respond once every *14* days. (They then have 14 days to respond.) https://t.co/oCAa4pFFNu — PolitiTweet.org

Nate Silver @NateSilver538

Here's something odd about the USC poll that makes it hard to interpret. The poll consists of a panel of people who are periodically asked about their vote. Nothing inherently wrong with this; can be a good way to detect changes in voter preferences. https://t.co/bmB61cZ8mX — PolitiTweet.org

Nate Silver @NateSilver538

I can't believe, after 2016, that people are still modeling the Electoral College as though each state is independent and giving Biden a 99.5% (?!?) chance of victory. State polling errors are highly correlated. This is bad math and is misinforming people. https://t.co/Jw1T8zGhX7 https://t.co/gByInNiCgE — PolitiTweet.org

Nate Silver @NateSilver538

@skepticalsports I think there's a utilitarian defense of it that it got people on the same page in the short run, and the short run mattered a lot and was pretty important, but I think that defense should concede that it created confusion later. — PolitiTweet.org

Nate Silver @NateSilver538

@skepticalsports It's still not clear to me whether it was intended to signal suppression even though it might have seemed to imply mitigation, or whether it was indeed a call for mitigation as opposed to suppression. Probably a mix of both depending on who you ask. — PolitiTweet.org

Nate Silver @NateSilver538

And in general, imputing the national popular vote from state polls is actually a bit more accurate than national polls on their own. It's a lot more work to do it, but sometimes it's worth it, e.g. state polls correctly implied a larger Obama win in 2012 than national polls did. — PolitiTweet.org

Nate Silver @NateSilver538

There's a lot going on with the forecast. It's regressed toward a prior that assumes the race will tighten, but the prior is weighted less over time. It has a subtle convention bounce adjustment. Still, it relies heavily on state polls, and it's *not* clear those have tightened. — PolitiTweet.org

Nate Silver @NateSilver538

But again, though, it's not entirely clear that the race is tightening. Our popular vote *forecast* (a forecast, not a polling average or snapshot) is holding pretty steady at Biden +6.7. https://t.co/ajG88SznSA https://t.co/mCzYIA0BY8 — PolitiTweet.org

Nate Silver @NateSilver538

Do I think that's what's going on here? To be honest, *probably* not; I think its more likely Trump is being helped a bit by (slightly) improving COVID and economic conditions, plus it's hard to sustain a 8-9 point lead in a polarized climate. But, it's hard to say. — PolitiTweet.org

Nate Silver @NateSilver538

FWIW, convention bounces can persist a bit longer than people might assume. Our empirical analysis says it can affect polls for ~3 weeks from the final date of the convention. We're still (barely) within that period, and keep in mind that some polls are released with a lag. — PolitiTweet.org

Nate Silver @NateSilver538

One other thing: Especially with ABC/WaPo and NYT Upshot/Siena leaning heavily into state polls this year, you do see higher quality in the state polls on average. That's where most of the live-caller data is going. — PolitiTweet.org

Nate Silver @NateSilver538

The twist here is that PA and FL are probably the two most important states so they're really hoping to keep Trump's hopes afloat. Still, it's pretty hard to tell if it's tightening or not, and also if there are distinctive regional patterns or it's just kind of random. — PolitiTweet.org

Nate Silver @NateSilver538

More on that state vs. national poll discrepancy. Biden's lead in our national polling average has shrunk from +8.4 pre-conventions to +6.9 now. But in swing states, he's basically unchanged on average, with gains in AZ/MN/etc. offset by declines in PA/FL/etc. https://t.co/LceSE7OC0K — PolitiTweet.org

Nate Silver @NateSilver538

There are a handful of eye-popping Biden results in state polls in recent days (+21 in Maine, +16 in Minnesota, +10 in Arizona, +10 in Wisconsin) and yes those are probably outliers but it's good to see those because it means that pollsters aren't herding. — PolitiTweet.org