Nate Silver

Nate Silver @NateSilver538

If we run the 2020 version of our model on 2016, it gives Clinton a 70% chance of winning the Electoral College, vs. 71%/72% in our actual final 2016 model (polls-only/polls-plus). So adding in 2016 data doesn't change much; it was a pretty normal-sized polling error. — PolitiTweet.org

Nate Silver @NateSilver538

The good reasons to be cautious about what will happen in November involve (i) coronavirus and all the other news and (ii) the longer-term history of polling 100 days out not being all that accurate. The bad reasons involve that Trump won as a mere 2:1 or 3:1 underdog in 2016. — PolitiTweet.org

Nate Silver @NateSilver538

RT @aedwardslevy: Because this is a good use of time on Saturday night, I want to just pull out this 2016 post-mortem from the pollsters' p… — PolitiTweet.org

Nate Silver @NateSilver538

Not much to say here. Saturdays are fairly average days; it's Sundays/Mondays that tend to feature lulls in reporting. Still, over 1000 deaths for the 5th consecutive day and the 7-day average is still climbing. — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking Newly reported deaths Today: 1,037 Yesterday: 1,178 One week ago (7/18): 872 Newly reported cases T: 65K Y: 75K 7/18: 65K Newly reported tests T: 798K Y: 930K 7/18: 762K Positive test rate T: 8.2% Y: 8.1% 7/18: 8.6% — PolitiTweet.org

Nate Silver @NateSilver538

If I ran a sports franchise I'd want to trade present value for future assets on the theory that I'm more likely to be able to maximize revenue from my team in 1-2+ years when my stadium/arena is fully open to fans again. — PolitiTweet.org

Nate Silver @NateSilver538

@mattyglesias Yeah. Like, the people who select to live in the most far-flung and/or rural parts of what's technically the NYC metro area are really selecting *against* NYC (but possibly remaining in the metro because of work or family ties). And they tend to be quite Republican. — PolitiTweet.org

Nate Silver @NateSilver538

One thing we definitely find is that medium-density precincts on the outskirts of a major city are fairly Republican, whereas medium-density precincts at the core of their metro areas (think AA or AAA cities in the sense of minor league baseball) are quite Democratic. — PolitiTweet.org

Nate Silver @NateSilver538

RT @kjhealy: If you're participating in "I have an x joke, but—", here's my helpful and objective field guide to social scientists. https:/… — PolitiTweet.org

Nate Silver @NateSilver538

@ForecasterEnten Larry wouldn't change a thing but Harri would be hugely different. People underrate the importance of how names are spelled IMO. — PolitiTweet.org

Nate Silver @NateSilver538

Otherwise, this is the not-very-good pattern that we're used to. Deaths up week-over-week, other numbers mostly flat. — PolitiTweet.org

Nate Silver @NateSilver538

Well, the news wasn't *all* bad. We did set a record for the number of new tests conducted in a day (930,000!) although it's not clear if that reflects an actual increase in testing capacity or if states just happened to do a lot of reporting today. — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking: Newly reported deaths Today: 1,178 Yesterday: 1,039 One week ago (7/17): 951 Newly reported cases T: 75K Y: 71K 7/17: 77K Newly reported tests T: 930K* Y: 774K 7/17: 837K Positive test rate T: 8.1% Y: 9.2% 7/17: 9.2% * record — PolitiTweet.org

Nate Silver @NateSilver538

Maybe these aren't necessarily contradictory. In practice, the improvement could be fairly incremental and linear. But either the arrival of a vaccine, or its failure to arrive, could lead to a nonlinear repricing of risk. — PolitiTweet.org

Nate Silver @NateSilver538

The second, somewhat contradictory answer is that we shouldn't look at a vaccine as an inflection point but as part of a series of incremental improvements on a number of fronts that will take place (probably with speed bumps) over the next say two years. — PolitiTweet.org

Nate Silver @NateSilver538

This is a good piece and I'm sort of torn between two answers. One is that the winter could be very tough, and if there isn't a vaccine that's fairly widely available by the spring, people are going to be ready to give up on most remaining lockdowns/distancing at that point. — PolitiTweet.org

Nate Silver @NateSilver538

@WesPegden Yeah, it's quite rational in some sense! Once every X number of elections, the polls wind up having a Democratic-leaning bias and you can dine out on those elections for a long time if you play your cards right. — PolitiTweet.org

Nate Silver @NateSilver538

https://t.co/CSL1oeowvl — PolitiTweet.org

Nate Silver @NateSilver538

Like imagine there was a pollster that nearly always had a Democratic bias over many elections and this year they doubled down and constantly published polls showing Biden up 20 points in the popular vote. That's basically what's happening here. — PolitiTweet.org

Nate Silver @NateSilver538

It's sort of amusing (and has some parallels to the media ecosystem overall) how Rasmussen Reports just brazenly doesn't give a f*ck about publishing outlier polls showing Trump at 49% approval despite a long history of overrating how well Republicans would do. https://t.co/MiZ6LhfioV — PolitiTweet.org

Nate Silver @NateSilver538

https://t.co/2cJbc2Dysh — PolitiTweet.org

Nate Silver @NateSilver538

Pretty sure I've reached the end of the internet. https://t.co/EYNGf7fxcQ — PolitiTweet.org

Nate Silver @NateSilver538

My basic priors are that online polls sometimes dampen real swings (by overly modeling their data) and that live polls sometime exaggerate them (because of non-response bias), so taking a mix probably isn't a bad idea. — PolitiTweet.org

Nate Silver @NateSilver538

As others have pointed out, there's also a pretty big split right now between live-caller polls, which show huge margins for Biden, and online polls, which show the race tightening a bit. — PolitiTweet.org

Nate Silver @NateSilver538

With Fox News's poll out tonight, we finally have enough Minnesota polls to show an average, and it shows Biden +11. So, no indication the protests have helped Trump out at all there. https://t.co/WfszY2xOvB https://t.co/5TmaRQvzy6 — PolitiTweet.org

Nate Silver @NateSilver538

More of the same. Deaths up week-over-week, and third day in a row over 1,000. Other metrics flat. In fact, the positive test rate was on the higher side today, though it's pretty consistently been somewhere in the 8-9% range. — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking: Newly reported deaths Today: 1,039 Yesterday: 1,126 One week ago (7/16): 974 Newly reported cases T: 71K Y: 70K 7/16: 71K Newly reported tests T: 774K Y: 796K 7/16: 828K Positive test rate T: 9.2% Y: 8.8% 7/16: 8.6% — PolitiTweet.org

Nate Silver @NateSilver538

I just want to point out that a *lot* has gone wrong since we did this. — PolitiTweet.org

Nate Silver @NateSilver538

This is why modeling is tricky, by the way. Are I comfortable assuming that COVID has zero effect on turnout patterns? (In ways that could increase polling error?) Not really. Is there a good way to model it empirically? Also, not really; no great precedents for COVID. — PolitiTweet.org

Nate Silver @NateSilver538

I haven't seen a study on this. But there are roughly a million unpredictable effects like this that COVID could have, perhaps in a quirky way state by state, and I keep going back and forth on whether to try to account for that somehow in our model ... leaning "yes". — PolitiTweet.org