Nate SilverVerified account

But this isn't rocket science. Senate seats last for 6 years. Redistricting lasts for 10 years. The change in the media narrative because Biden happened to have a marginally prettier-looking map might last for a few weeks.

I am also convinced that the difference between, say, 52 and 53 Democratic senate seats could matter. And state legislative races could matter a lot given the impact on redistricting.

...but the states that could flip in a somewhat close election (say, North Carolina) are probably places where you want to be invested anyway, just because it's hard to guess exactly what the tipping-point states will be ahead of time.

I am more convinced that the difference between a *very* close election that isn't called for several days/weeks and a *somewhat* close election (say, a 5 point Biden popular vote margin called within 24 hours) could matter...

I am just not convinced that, if Biden wins the election by 8 points or something, whether he happens to win say Ohio or Georgia as a part of that would really make much difference in the long run.https://twitter.com/jmartNYT/status/1281953272522330115 …

And keep in mind that when the numbers were rising in the spring, at least we knew that lockdowns meant they would probably eventually start to come back down. Not the case this time.

The comparison to last Saturday is somewhat misleading since it was a holiday weekend. But the number of deaths *two* Saturdays ago was 509 and we are also way up from that. Just no good news here, and the positive test rate keeps climbing.

US daily numbers via @COVID19Tracking Newly reported deaths Today: 757 Yesterday: 854 One week ago (7/4): 306 Newly reported cases T: 63K Y: 67K 7/4: 52K Newly reported tests T: 634K Y: 823K 7/4: 642K Positive test rate T: 9.9% Y: 8.1% 7/4: 8.1%

We did see a record number of test results reported, which may reflect a backlog as it's now taking a while to process tests given high demand for testing. That backlog could make it harder to know precisely whether things are getting better or worse at any given moment.

I mean... kinda speaks for itself. Looks like deaths are on the rise again and it's not just lingering effects from the holiday weekend. That's what's going to happen if you go from 20K cases a day to 60K+, even if a lower *share* of cases result in deaths than in the spring.

US daily numbers via @COVID19Tracking: Newly reported deaths Today: 854 Yesterday: 867 One week ago (7/3): 624 Newly reported cases T: 67K* Y: 59K 7/3: 57K Newly reported tests T: 847K* Y: 637K 7/3: 719K Positive test rate T: 7.9% Y: 9.2% 7/3: 8.0% * record

But the media only considers polls "wrong" if one candidate is ahead and the other one wins, even if the miss is within the margin of error (as it was, more or less, in 2016). If Biden's ahead by 6 in the final polls and wins by 13, nobody will ask what went wrong with the polls.

There had been years where polls were fairly far off the mark. In 1980, Reagan was just barely ahead in final polls but won in a landslide. In 1996, Clinton underperformed his polls but still won convincingly. State polls were a bit off in 2000.

It's tempting to say "people only remember the last election". But one thing about 2016 is there hadn't been a recent presidential election where the polls had "called" the winner wrong* (you have to go back to 1948), so it'll probably stick in people's heads for awhile.https://twitter.com/databyler/status/1281681011038801921 …

Also, while the uncertainty is undoubtedly wide enough to include some Trump wins, it's very much bidirectional; it's not hard to imagine scenarios where things get worse for Trump, which could lead to some very large (e.g. low-to-mid teens in the popular vote) Biden wins.

Uncertainty around the outcome of the election is high, but it's high for different reasons than at the end of 2016. It's high mostly because there's a fairly long way to go and there's a lot of weighty news that could (at least in theory) swing things in either direction.

Since data folks are accused of failing to account for real-world context: reporters should look at the real world when writing poll stories like this! More than 130K Americans are dead of a pandemic the president said would go away. Not hard to explain why he's down by 9 points.https://twitter.com/NateSilver538/status/1281658769911566337 …

But one of the basic things that people screwed up in 2016 is that the election was *pretty freakin' close* at various points, including at the end of the campaign. Clinton went into Election Day with a 2-3 point lead in the average tipping point state. That's highly competitive.

If polls were off by *twice* as much as in 2016, then it starts to matter a bit more where the errors occurred and the election gets competitive, though Biden probably a narrow favorite.