I feel kinda sad for all the people who look at this and think there aren't enough rural voters:
Census-defined rural PA: 33% of raw sample; 33% of PA registered voters
Census-defined rural OH: 37% of raw sample; 37% of OH registered voters.
We weight our samples by census block group density.
The lowest quartile of census tract density represents 27% of the raw sample in OH and 26% in PA, so we slightly down-weight folks from less dense tracts.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
One fascinating thing is that most people look at these maps and think there are too many people in the cities.
In both states, the raw sample is slightly too rural (and weighted appropriately)
There are lots of things in polling that are hard to get right. The *number* of voters in rural areas is not hard to get right. The number of rural RVs, for instance, is just a cold hard fact. We can make our polls match it.
As you can see, we have the... latitude and longitude on our respondents. That's an extremely powerful tool for getting urban-rural splits right.
Over all, I do think the NYT/Siena results in OH/NV and the Marquette Law results for WI are *relatively* good for the president. They're more like the polls we saw before the debate than the Biden>10 type numbers we've seen a lot of lately
Now, I think these are two pretty good pollsters. But we're at the point where the 'good' news for the president is basically the old average, and that's generally a sign things have shifted.
FWIW, I don't see much evidence of nonresponse bias in our polling, or, frankly, even in the polls that do show Biden doing super well. That Biden+12 in PA was D+1 among RVs, after all, and maybe even R+1 among LVs, like our PA poll, @PollsterPatrick ?
Joe Biden leads Donald Trump in Ohio and Nevada, according to new Times/Siena polls.
Biden leads in Ohio by 1 point, 45 to 44 percent.
Biden leads in Nevada by 6 points, 48 to 42 percent. nytimes.com/2020/10/07/us/…
The polls began on Sat., after the president's diagnosis. FWIW, the president's best interviews were yesterday, after he was released from the hospital, controlling for demog. That said, it's always tough to judge one night--let alone the last night, when we're finishing quotas
Unlike in our FL/PA release, when we had AZ already started heading into the last night of PA/FL interviews, we didn't have another poll under way on Monday to help make this kind of analysis more robust
I mean, by now you should probably know that you all have very little ability to predict our results based on this information, but:
OH: R+3 party ID, Trump+8 in recalled '16 vote
NV: D+4 party ID, Clinton+2 in recalled '16 vote
Joe Biden leads Donald Trump in Arizona, 49 to 41 percent, in a new Times/Siena survey.
Democrat Mark Kelly leads Republican Martha McSally, 50 to 39 percent, in the US Senate race nytimes.com/2020/10/05/us/…
This is not the first time we've had some better-than-average numbers for the Democrats in Arizona, so I looked a bit under the hood at something potentially material: the number of new voters. nytimes.com/live/2020/pres…
In our poll, voters without a record of voting in 2016 make up 32% of the electorate, and they break overwhelmingly for Biden--who leads by 2 pts among likely 2016 voters.
Joe Biden leads Donald Trump in Pennsylvania and Florida, according to new Times/Siena polls taken after the first debate.
Biden leads in Pennsylvania, 49 to 42 percent among likely voters. He leads in Florida, 47 to 42. nytimes.com/2020/10/03/ups…
In an analysis of interviews conducted yesterday, including in Arizona (where a poll is ongoing), there was modest evidence of a shift in Joe Biden's direction after Trump's COVID diagnosis. That said, one day of interviews is not nearly enough to reach any firm conclusions
The shift, though, would be material if confirmed and it was statistically significant, controlling for the demographic and political variables used in weighting. That said, it's still just one day of interviews. We'll have to wait and see here.