Nate Silver

Nate Silver @NateSilver538

RT @galendruke: It is high season for VP speculation https://t.co/L9bu0HYQ5v — PolitiTweet.org

Nate Silver @NateSilver538

Our model tries account for these correlations based on geographic and demographic factors. e.g., in some random number of simulations, Biden will do better than polls show among Hispanics or worse in the Midwest. So we'll add COVID to the list of factors that can create them. — PolitiTweet.org

Nate Silver @NateSilver538

i.e. Biden could do "surprisingly" well in Arizona, which has been hard-hit, but underperform in New Mexico, which hasn't been. Or he could do well in both the Northeast and the South (hardest-hit regions) but not gain as much in the Midwest and the West. — PolitiTweet.org

Nate Silver @NateSilver538

One other factor to consider in how COVID could affect election forecasts is that variation in state COVID rates could conceivably be a factor in support (or lack thereof) for Trump, which could create some unusual correlations. https://t.co/MVZOD5xlnn — PolitiTweet.org

Nate Silver @NateSilver538

One other factor to consider in how COVID could affect election forecasts is that variation in state COVID rates could conceivably be a factor in support for Trump, which could create some unusual correlations: — PolitiTweet.org

Nate Silver @NateSilver538

The other numbers are actually mildly encouraging. Two days in a row with the positive test rate at 7% after weeks at 8-9%. And the lowest new cases total since July 7, though still very high. Mondays are often slow, though, so the midweek numbers will tell the story. — PolitiTweet.org

Nate Silver @NateSilver538

† The death total today included a one-off adjustment from Texas, counting deaths that may have occurred some time ago. See below for more detail. Still, deaths increased week over week even if you don't count those toward today's total. https://t.co/t3sx74WYqM — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking: Newly reported deaths Today: 428 + 631† Yesterday: 558 One week ago (7/20): 362 Newly reported cases T: 55K Y: 62K 7/20: 57K Newly reported tests T: 761K Y: 856K 7/20: 711K Positive test rate T: 7.2% Y: 7.2% 7/20: 8.0% — PolitiTweet.org

Nate Silver @NateSilver538

RT @wiederkehra: So I promised more behind-the-scenes of the @FiveThirtyEight 2020 forecast. Today I'll walk you through the development of… — PolitiTweet.org

Nate Silver @NateSilver538

My point is more that I wouldn't be wanting to use an overly precise or deterministic prior when there is a ton of economic uncertainty (we'd guess ~3x higher than normal) and different variables say different things about the state of the economy. — PolitiTweet.org

Nate Silver @NateSilver538

So, the *prior* shifts by ~5 points, but our model doesn't actually weight the prior that heavily because we think the relationship the economy and incumbent performance isn't actually that strong. A 5-point shift in the prior might imply a 1-2 point shift in the polls. — PolitiTweet.org

Nate Silver @NateSilver538

I'd also note that our economic index uses disposable income as one of its 6 variables and Trump/the economic index is currently getting a big boost from transfer payments. If those lapse, may underperform projections. — PolitiTweet.org

Nate Silver @NateSilver538

I'm not sure that I personally buy the V-shaped recovery story at all, but that's what you get if you use surveys of professional forecasters + the SP&500 (yes, it's predictive of other economic variables like payrolls) + a little bit of mean reversion. — PolitiTweet.org

Nate Silver @NateSilver538

One thing about models that use economic data as part of their election forecast is the economy is expected to materially improve by November. We have Trump as a 9-10 point popular vote underdog based on current conditions, but only 4-5 points using projected Nov. conditions. — PolitiTweet.org

Nate Silver @NateSilver538

@jodyavirgan Only one team has had to be quarantined so far! At this rate we'll get at least 10 or 12 games in before they cancel. — PolitiTweet.org

Nate Silver @NateSilver538

For people wondering what our model actually says about 2020: I don't actually know yet! The programming is mostly done, but processing 2020 is the last step. Making design choices *without* knowing how it'll affect the current forecast is a helpful way to stay at arm's-length. — PolitiTweet.org

Nate Silver @NateSilver538

@TheStalwart NYC is instructive because it counts probable deaths, which most places don't (although an increasing number do). In mid-April, probable deaths were 36% of total (confirmed+probable) deaths in NYC. Now that has fallen to 20%. — PolitiTweet.org

Nate Silver @NateSilver538

@TheStalwart There were likely far more unrecorded deaths (both as a proportion of all deaths and in absolute numbers) in March/April than there are now. — PolitiTweet.org

Nate Silver @NateSilver538

Since most swing states tend to have diverse-ish demographics that resemble the country in microcosm (though there are some exceptions like New Hampshire) that tends to make them easier to poll. This is also why national polls are usually more accurate than state polls. — PolitiTweet.org

Nate Silver @NateSilver538

For instance, in Utah if you don't get the right number of Mormons or a representative sample of Mormons, your poll is going to be screwed. But that won't affect your numbers as much in, say, Ohio. — PolitiTweet.org

Nate Silver @NateSilver538

Not sure if people realize this, but empirically, states with diverse demographics are easier to poll than states where a large fraction of the population belong to just one or two groups. — PolitiTweet.org

Nate Silver @NateSilver538

Meanwhile, his approval rating in polls among registered/likely voters really isn't any better and in fact is the lowest it has been since Jan. 3, 2018. https://t.co/x0RAzFPReK — PolitiTweet.org

Nate Silver @NateSilver538

Trump's approval rating drops below 40% (albeit just barely, to 39.9%) for the first time since the shutdown. https://t.co/Vfmzd6B2ps https://t.co/J1rdODq4QU — PolitiTweet.org

Nate Silver @NateSilver538

I am more skeptical that Trump would get credit for it, though, especially because the cumulative pain from COVID & lockdowns may still be getting worse by Nov. even if cases/deaths are declining. Possibly also the economic fallout, especially if Republicans screw up on stimulus. — PolitiTweet.org

Nate Silver @NateSilver538

I do think the punditry slightly underrates the chance of 1a) (i.e. the COVID situation getting better) just because it's been pretty hard to predict then the ebbs and flows in COVID spread and the COVID models have wide ranges for what'll happen by November 3. — PolitiTweet.org

Nate Silver @NateSilver538

"Something else" might include a fairly long menu of options, range from a Biden gaffe to some military entanglement to a Supreme Court vacancy to simply a large polling error. Of course, there are also wild cards/"something elses" that could make Trump's position even worse. — PolitiTweet.org

Nate Silver @NateSilver538

Yeah, my general read that to make the election competitive, Trump probably needs (1a) the coronavirus situation to get objectively better by November; (1b) to receive *some* degree of credit for it and (2) "something else". — PolitiTweet.org

Nate Silver @NateSilver538

@BenjySarlin @NickRiccardi @joshtpm The problem is that, although Europe had much stricter lockdowns than the US in March/April, it's not clear to me that they have stricter ones *now*, *or* that they took longer to scale back. The timelines and scope of relaxations were fairly similar from what research I've done. — PolitiTweet.org

Nate Silver @NateSilver538

@BenjySarlin @NickRiccardi @joshtpm But the hot parts of Europe still generally have a culture of hanging out outdoors, and to the extent they are indoors, not having air conditioning. Warm weather per se may not be a problem for COVID transmission, and may in fact be helpful. — PolitiTweet.org

Nate Silver @NateSilver538

@BenjySarlin @joshtpm The questions IMO for both Europe and the US Northeast are what happens when cooler weather forces people indoors, if immunity is short-lived (though I think that concern is generally a bit overblown), or fatigue erodes compliance with even moderate distancing. — PolitiTweet.org