Nate Silver

Nate Silver @NateSilver538

Empirically, these are about equally likely. That is, the direction of polling bias in one election year is not at all predictive of the direction of bias in the next year. You could have a R bias and then a big D bias four years later, for instance. https://t.co/ees0U6j4L7 — PolitiTweet.org

Nate Silver @NateSilver538

Put another way, after a polling "miss", there is the risk of under-compensation (you change nothing make the same mistakes as last time) or of overcompensation (you change things that were working fine and introduce bias in the other direction). — PolitiTweet.org

Nate Silver @NateSilver538

Keep in mind that pollsters actually *underestimated* Clinton in some states (e.g. CA, AZ) in 2016. And that there was little overall bias in 2018, with errors running in different directions in different states. — PolitiTweet.org

Nate Silver @NateSilver538

Post 2016, it's likely that pollsters will spend more time scrutinizing practices that resulted in a D bias, while spending less time on those (say, not reaching enough Hispanics) that resulted in a GOP bias. — PolitiTweet.org

Nate Silver @NateSilver538

I *do* think pollsters should weight by education and it's probably correct that failure to weight by education will cause a D bias. But, there are a lot of debatable practices, which can/will produce biases in both directions. — PolitiTweet.org

Nate Silver @NateSilver538

I remain convinced that by far the best macro strategy is to take an average of all polls, but weighting toward higher-quality polls, rather dismissing/ignoring individual polls because they don't meet some particular methodological standard (e.g. not weighting by education). — PolitiTweet.org

Nate Silver @NateSilver538

Meanwhile, the positive test rate continues to decline. Over the past 7 days, it has been 6.2%, roughly half what the 7-day trailing average was on May 1 (12.1%). That may partly reflect more widespread availability of testing in the general population but still a good sign. — PolitiTweet.org

Nate Silver @NateSilver538

Tuesdays are often quite bad as weekend reporting backlogs are cleared. But the past two Tuesdays have shown some improvement (1,430 and 1,543) after several Tuesdays in a row with 2,000+ deaths. We'll see what tomorrow and Thursday (also days that can produce bad numbers) bring. — PolitiTweet.org

Nate Silver @NateSilver538

US daily numbers via @COVID19Tracking Newly reported deaths Today: 1,430 Yesterday: 786 One week ago (5/12): 1,543 Newly reported cases T: 21K Y: 20K 5/12: 21K Newly reported tests T: 399K Y: 338K 5/12: 305K Positive test rate T: 5% Y: 6% 5/12: 7% — PolitiTweet.org

Nate Silver @NateSilver538

@asymmetricinfo @JamesSurowiecki @chrislhayes @SeanTrende @bwreed @guypbenson @RameshPonnuru I think that intuition is right, though it depends on how we're specifying the question. Interventions (masks, social distancing) can/should dampen differences in R that result from density and environmental factors, as should whatever reduction in susceptibility. — PolitiTweet.org

Nate Silver @NateSilver538

@asymmetricinfo @JamesSurowiecki @chrislhayes @SeanTrende @bwreed @guypbenson @RameshPonnuru I mean, it depends on what you mean by huge. If one city has an initial cluster on Feb. 20 with an R of 2.2, and another has an initial cluster on March 1 with an R of 1.7, those will produce hugely different results after a couple weeks (roughly 10-fold by March 15). — PolitiTweet.org

Nate Silver @NateSilver538

@asymmetricinfo @JamesSurowiecki @chrislhayes @SeanTrende @bwreed @guypbenson @RameshPonnuru It is colder, though (especially in March). And it's poorer. And undoubtedly there is a lot of stochasticity, especially in the early phases where there are varying numbers of early introductions from place to place and nobody is taking any precautions. — PolitiTweet.org

Nate Silver @NateSilver538

@chrislhayes @SeanTrende @bwreed @guypbenson @RameshPonnuru I think part of the problem is people are conditioned to think about a NYC or Lombardy type explosion in cases when they see headlines about "spikes", etc., whereas populationwide exponential growth that fast may turn out only to occur under a particular set of circumstances. — PolitiTweet.org

Nate Silver @NateSilver538

@chrislhayes @SeanTrende @guypbenson @RameshPonnuru I don't know, given that this is likely to continue to be the dominant news story for months (or years!), I think Georgia ought to cause a fair amount of self-reflection, because it was clear at the time some of the analysis was incautious and was risking a potential own-goal. — PolitiTweet.org

Nate Silver @NateSilver538

The other states are generally re-opening also, and not on terribly different schedules than GA/FL/TX. They've just received less media fanfare while doing so, and their governors are more popular. — PolitiTweet.org

Nate Silver @NateSilver538

One plausible theory is if there's a big media brouhaha over re-opening, it may actually make people more cautious: * They may perceive more risk * They may want to set a good example if their state is scrutinized * Dems in the states may view staying home as a partisan signal — PolitiTweet.org

Nate Silver @NateSilver538

GA, FL & TX have probably been the most focused-upon states for opening early, but if you look at mobility data, they're on the low end compared to neighbors. https://t.co/32iSFYgSGd LA -29% *FL -27%* *TX -14%* NM -8% TN -3% *GA -2%* AR -1% NC +1% MS +2% OK +5% AL +10% SC +21% — PolitiTweet.org

Nate Silver @NateSilver538

Universities announcing plans to bring students back in the fall (not all of them, but an increasing number) seems like a pretty important inflection point. — PolitiTweet.org

Nate Silver @NateSilver538

Although the situation undoubtedly varies from state to state, so check local guidance, this has actually become more common. There is now excess testing capacity in some states that isn't being fully utilized. https://t.co/EVCQSX7N9x — PolitiTweet.org

Nate Silver @NateSilver538

Testing capacity has increased a lot in New York (both city and state). First it was Cuomo and now it's AOC urging more people to get tested. — PolitiTweet.org

Nate Silver @NateSilver538

RT @StartHereABC: Taking a closer look, @NateSilver538 joins this morning’s conversation to examine #COVID19 trends - LISTEN: https://t.co/… — PolitiTweet.org

Nate Silver @NateSilver538

New York City now with just a 4.6% positive test rate in today's data, which is actually a bit *lower* than the US as a whole (5.8% yesterday). https://t.co/1w2VrzC4FG — PolitiTweet.org

Nate Silver @NateSilver538

The question is whether 6 feet is a good rule-of-thumb that people can remember and coordinate around, are generally willing to abide by, and which considerably reduces risk in the vast majority of circumstances. If it's not, then people should propose a better rule-of-thumb. — PolitiTweet.org

Nate Silver @NateSilver538

Yeah, the "ackshawlly, 6 feet is an arbitrary distance" folks aren't super helpful. Of course it's arbitrary. You don't magically go from RISKY at 5.99' to SAFE at 6.01'. Instead risk presumably declines as some nonlinear function of distance X exposure. — PolitiTweet.org

Nate Silver @NateSilver538

Importantly also, there's less uncertainty in the very near term. That is, it's unlikely there's a huge spike or rapid decline by the end of this month. It's the accumulation over time. Deaths growing at 10% a week vs. shrinking at 10% makes a big difference over a few months. — PolitiTweet.org

Nate Silver @NateSilver538

But look at the uncertainty! Many unknowns. It's not hard to imagine some parameters where we're down to "just" 300 deaths a day (or fewer) by mid-August! Or have bounced back to 3,000 (or more)! IMO this reflects a well-designed model that accounts for real-world uncertainty. — PolitiTweet.org

Nate Silver @NateSilver538

These @youyanggu projections are interesting. If his model had to make a modal projection, it would be that the US will plateau somewhere near the recent total of ~1,300 deaths per day as restrictions are relaxed to the point where we get an R of ~1. https://t.co/Qxu8rSPoV1 https://t.co/7xi655aCxm — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris Please do, that's what the data is there for! — PolitiTweet.org

Nate Silver @NateSilver538

RT @RichieFed: Karen’s husband, Kevin, is very rude! 🤦‍♂️ https://t.co/Y27nLBp7sG — PolitiTweet.org

Nate Silver @NateSilver538

@gelliottmorris That's already accounted for the ratings, so pollsters are treated more forgivingly if they poll earlier. — PolitiTweet.org