Nate Silver

Nate Silver @NateSilver538

@gelliottmorris @DanRosenheck @SethS_D Both priors pointed toward a popular vote tossup at this point and both received similar (and low) weight. We do not necessarily agree that Biden's prior should be better given that Trump is an elected incumbent + economy is improving & forecasted to improve more. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D Or you can compare individual states. We had Clinton at +5.4 in PA in at this point in 2016 and 81% to win, whereas this year we have Biden +5.0 in PA and 75% to win. So it's pretty darn similar. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D Right. As you can see in the follow-up tweet, her margins in the tipping points in polls-plus were pretty similar to where we have Biden projected this year, if not perhaps a half-point or a point better. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D Actually, let me amend that ... that's in polls-only. In polls-plus (more comparable to the 2020 version) she was at +5 in PA, +8 in WI, +8 in MI, +4 in NV, though only +2 in FL. But overall, that's pretty similar to if not a bit better than where we have Biden this year. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D On Aug. 14, which was close to her high-water mark, we had Clinton projected to win by 9 points in PA, 11 in MI & WI, 6 in FL, 7 in NV. So if anything her margins were a bit bigger in the tipping-point states. https://t.co/ZH2fUH7E8i — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D I'm not saying they completely cancel out. But, it's a fairly modest effect net-net. I'm also not sure what you're trying to prove. We think it's a big mistake not to consider the effects that COVID and the huge economic disruption has on the election. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D We haven't "added a whole bunch of uncertainty". The world has added it. And it's our job to look at reality and figure out how we can measure it. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D Also, I don't understand why there's any theoretical basis for ignoring COVID! Models should look at the real world and figure out how to reflect it! If there had been a COVID-15, we'd certainly have considered that in '16. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D We have Biden at +5 in our current tipping point state, Pennsylvania. So, that's decently similar. Also, there are uncertainties created by COVID and it is correct for a model to account for them because that's part of the real world! — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D I don't know. But probably fairly similar. Our uncertainty index includes several factors that reduce uncertainty that we hand't considered in earlier years (i.e. this is the first time we're looking at polarization), as well as others that increase it. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @gelliottmorris @SethS_D We didn't make a lot of changes. It's equivalent to like a new version of the EA Madden franchise. 90% the same product. The 2020 version would have produced a nearly identical forecast in 2016 as our 2016 model. It would have had Clinton at 70% on Election Day, for instance. — PolitiTweet.org

Nate Silver @NateSilver538

While people need to be are of the fact that Trump still has decent winning chances (27% in our model), there's also about a 33% chance that he suffers the biggest popular vote loss for an incumbent since Hoover (exceeding Carter in 1980). The uncertainty runs in both directions. — PolitiTweet.org

Nate Silver @NateSilver538

RT @Tonyhkchow: Our first 2020 Election Forecast update with @NateSilver538! We'll be doing more and more of these as Election Day gets cl… — PolitiTweet.org

Nate Silver @NateSilver538

In other words, additional vigilance by voters, postal-service workers, elected officials and the media likely goes a fairly long way, but it could be an extremely dangerous situation without that vigilance. — PolitiTweet.org

Nate Silver @NateSilver538

I'm not sure whether people are too worried or not worried enough about the post office stuff. But it's the sort of thing where the more that people are worried, the less need there may need to be to worry, and vice versa, if that makes any sense. — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris I don't agree that the specification differences are small, especially if you look at the popular vote, which is a more straightforward calculation. — PolitiTweet.org

Nate Silver @NateSilver538

In that sense, the best tip-off that the forecasts are different is they have Biden at 97% to win the popular vote, and have sometimes been as high as 98% (and 99% before they revised their model) whereas we are at 82%. You're getting up to 97-98%, you're getting VERY confident. — PolitiTweet.org

Nate Silver @NateSilver538

Also keep in mind that these probabilities are drawn from distributions that stretch out on the tails. It doesn't take a lot to go from say 50% to 54% (so those forecasts are ~identical) but it DOES take a lot to go from 95% to 99% (so those are quite different) https://t.co/HbSJcpbW0Y — PolitiTweet.org

Nate Silver @NateSilver538

Maybe it's my background as a poker player, but distinguishing between a 28% chance and a 11% chance ... you'd be thrilled to get your money in with that sort of edge. Elections are different, and there's much higher model specification uncertainty, but it's still fairly big. — PolitiTweet.org

Nate Silver @NateSilver538

Some stuff I agree with here, but I *don't* agree with the idea that an 89% forecast (Economist) and a 72% forecast (538) are close. Maybe easier to see if you compare a 28% chance of Trump winning to 11%; we think a Trump win is 2-3x as likely. https://t.co/ptN9d6z5P5 — PolitiTweet.org

Nate Silver @NateSilver538

In other news, we've got a very big project up summarizing early voting options, mail voting options, and key voting deadlines in each state, a lot of which have changed during the pandemic. Check it out! https://t.co/NuzR3nYbdX — PolitiTweet.org

Nate Silver @NateSilver538

Don't get me wrong, I love RCP, but I do think our averages do a better job of picking up real vs. fake movement. Over roughly the same time period, they have Biden moving from +9.3 to +6.4 and now back to +7.7, so about 3x as much volatility. — PolitiTweet.org

Nate Silver @NateSilver538

A flurry of new national polls haven't really moved our numbers much. Biden +8.4. By comparison, he was up +7.8 points a weeks ago and +8.3 points two weeks ago. https://t.co/cy51vc5isJ https://t.co/c1UxsZbvll — PolitiTweet.org

Nate Silver @NateSilver538

RT @juruwolfe: States are trying to move quickly to ensure the pandemic doesn't keep people from voting. But that can make it difficult to… — PolitiTweet.org

Nate Silver @NateSilver538

Someone saying "I'm making very conservative assumptions" is almost always a tip-off that they are in fact making aggressive or bold assumptions. — PolitiTweet.org

Nate Silver @NateSilver538

Today's numbers are more in line with recent trends, where things are getting a bit better, but only V E R Y S L O W L Y and from a very high peak. — PolitiTweet.org

Nate Silver @NateSilver538

* I missed yesterday's report, when there was a discouragingly high number of deaths, but also a lot of weird reporting issues ... see here for more. https://t.co/kTv9H9J6XG — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking: Newly reported deaths Today: 1,163 Yesterday: 1,503 One week ago (8/7): 1,251 Newly reported cases T: 52K Y: 56K* 8/7: 54K Newly reported tests T: 881K Y: 658K* 8/7: 732K Positive test rate T: 5.9% Y: 8.5%* 8/7: 7.4% — PolitiTweet.org

Nate Silver @NateSilver538

Just imagine what the Cuomo-Harris-MayorPete-AOC primary will be like on Twitter. — PolitiTweet.org

Nate Silver @NateSilver538

This sort of sums up some election forecasting debates, too. I tend to come at models from more of the "epi" perspective here, i.e. we start with observed reality, and now we just have to come up with the right math for it. But ideally you have a synthesis of both of these views. — PolitiTweet.org