Nate Silver

Nate Silver @NateSilver538

In other words, some of the popularity boost for BLM in June may have reflected a reaction against Trump / negative partisanship. Voters give Trump TERRIBLE marks for his handling of race relations and people may flock to the opposite position of whatever he says on these issues. — PolitiTweet.org

Nate Silver @NateSilver538

An alternative inference is that BLM popularity is waning *because Trump is saying less about it*. His Twitter feed has made few references to "BLM"/"Black Lives Matter" over the past 6-8 weeks, and he's also used terms like "rioting" and "looting" much less often than in June. https://t.co/n5m2D4EOt5 — PolitiTweet.org

Nate Silver @NateSilver538

There's some polling out lately that shows a decline in support for BLM following a sharp increase in June. Some people have taken that to mean that Trump/Republican messaging on BLM is breaking thru. Plausible, but I'm not sure that's so straightforward. https://t.co/yKBz5SC1OC https://t.co/NqEWTecGAO — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris No. I don't think you're acting in good faith and I'm tired of engaging with you, and I made a mistake in doing so tonight. — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan @gelliottmorris And I don't think he's engaging in good faith when I pointed that out to him the last time this came up and he's just ignoring that very direct way to make a comparison. — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan @gelliottmorris Because as I've pointed out to George before, you can directly compare the *forecasted* margins of victory we show in each state to the win probabilities we showed and they show that the same margin translates to about the same win probabilities as in 2016. — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan @gelliottmorris There are just several layers of dishonestly, with the implication that our model incorporates far more uncertainty in 2020 than in 2016, which is not true. — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan @gelliottmorris He's also citing the win probability from the "polls-only" version of our forecast from 2016, rather than "polls-plus", which is the version that we're using in 2020. And he's not mentioning the fact that Biden is in the midst of his post-convention period, while Clinton wasn't. — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan @gelliottmorris We've made changes to our polling average algorithm since 2016. Since he looked up what our win probability was in 2016, it would have been more honest of him to also look up and cite what the 2016 version of our polling average was that the win probability was derived from. — PolitiTweet.org

Nate Silver @NateSilver538

RT @Nate_Cohn: There are some valid differences: --538 shifted from 'polls only' to 'polls plus' as default. polls plus in '16 was clinton… — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan @gelliottmorris I think it's dishonest for him to conflate our polling average with our forecast when he knows better, but FWIW, it is also not an honest representation of our polling average. We had Clinton up by 6 on this date in 2016, not by 4. https://t.co/sJBHxVdNSS — PolitiTweet.org

Nate Silver @NateSilver538

@BrendanNyhan He is talking about our forecast, because there are no probabilities associated with our polling average. But please post the strawman GIF again without bothering to make sure you got the details right. — PolitiTweet.org

Nate Silver @NateSilver538

This is pretty dishonest. Our current forecast is for Biden to win the popular vote by 6 points. 8 points (not 9) is his current lead in the polling average & the model expects the race to tighten. And our forecast has Biden winning by only 3-4 points in the tipping-point states. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @StatModeling I know the various ways that people can fuck up when building these forecasts, and I've fucked some of them up myself, so I can often look at a result and say "oops, there must be a bug there, or at least something that wasn't well considered", and more often than not, I'm right. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @StatModeling The reason I think those other numbers are wrong is because I've literally spent years digging into the weeds on the frankly fairly obscure set of problems you need to make election forecasts. Probably more time than anybody else. So my knowledge base is pretty strong. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @StatModeling I will say this, though. If I'd developed a model that showed Biden at **99% to win the popular vote in July**, I wouldn't have published it. Not because I was afraid of how it looked, but because that's obviously a misspecified model that doesn't pass the smell test. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @StatModeling I also think it's completely irresponsible (and will lead to worse forecasts) *not* to think about the effect that COVID-19 will have. How you specify it is another question. But the job of a forecast is to reflect the real world, and COVID is huge part of the real world today. — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @StatModeling When have I ever denied my own agency? What you're accusing me of is making choices based on incentives other than having the most accurate (in the sense of well-calibrated) model, so please be honest about what your accusation was. — PolitiTweet.org

Nate Silver @NateSilver538

There's a lot happening tonight, but Day 3 of our RNC liveblog is up: https://t.co/OxNH8cvyPk — PolitiTweet.org

Nate Silver @NateSilver538

RT @alexismadrigal: This is pretty big (and good) news. They are promising ~50 million~ tests a month by early Oct. For comparison, the cou… — PolitiTweet.org

Nate Silver @NateSilver538

There's no especially pithy conclusion here. It's just something people need to keep in the back of their heads and perhaps approach overly precise claims about shifts in the Black or Hispanic vote with a little bit of caution. — PolitiTweet.org

Nate Silver @NateSilver538

There's some polling evidence to suggest Trump is performing better than in 2016 with Black and Hispanic voters, especially younger Black and Hispanic voters. But polls have long had trouble reaching representative samples of Black and Hispanic voters (especially younger ones). — PolitiTweet.org

Nate Silver @NateSilver538

@DanRosenheck @StatModeling That's a pretty lame accusation, Dan. We don't look at what the results are going to be when we're working on the model. I was surprised it wasn't higher on Biden. — PolitiTweet.org

Nate Silver @NateSilver538

That's not to say there's no sequence of events that could play to Trump's benefit. But generally when something goes bad on your watch and you're the incumbent, it doesn't help you, and Trump often makes situations worse when inserting himself in them. — PolitiTweet.org

Nate Silver @NateSilver538

RT @perrybaconjr: The media plays an important role in our society, so strong criticism comes with the territory and sometimes that critici… — PolitiTweet.org

Nate Silver @NateSilver538

One thing I think pretty demonstrably *doesn't* help Trump is a chaotic news cycle, even though he sometimes seems inclined to cultivate one. It would be more the sense that the country was "in recovery" / getting back to normal that would likely help him. — PolitiTweet.org

Nate Silver @NateSilver538

Biden's at only 55-60% to win in prediction markets, but 85-90% in other statistical models, which seems like an insanely wide spread—and I don't think either number is super easy to defend—so I have to say I'm pretty happy that 538 model's falls somewhere in the middle. — PolitiTweet.org

Nate Silver @NateSilver538

@IChotiner But he probably is suffering a penalty for his COVID handling. People's default around the world seems to be to become more pro-incumbent under COVID. I do wonder what would happen if there were all the same policy failures but he came across more empathetically. — PolitiTweet.org

Nate Silver @NateSilver538

@IChotiner But he probably is suffering a penalty for his COVID handling. People's default around the world seems to be to become more pro-incumbent under COVID. I do wonder what would happen if there were all the same policy failures but he came across more emphatically. — PolitiTweet.org

Nate Silver @NateSilver538

While Trump will *probably* lose and *possibly* lose badly—indeed, he's *currently* on track to suffer the worst incumbent loss since Carter—I don't think people's priors should be terribly strong about what effects a pandemic might have on an incumbent's popularity. — PolitiTweet.org