2020 Poll Hype Thread

[Pollster]

So, who did this poll? Nv indy, but what pollster?

Mark Mellman - usually a Dem internal pollster, but apparently good friends with Ralston

[Pollster]

I expect both polls to show the top 4 in a virtual tie in the high teens/low twenties, with Pete maybe trailing the pack in Nevada.

Will be interesting to see if the surge of undecideds in Selzer's poll is shown here as well.

[Pollster]

Online polling when conducted properly is just as reliable as phone polling when conducted properly.

Methodology, not mode, is what matters.

[Pollster]

Progress Campaign teasing a Georgia poll for May 18

[Pollster]

Progress Campaign teasing a poll of Kenosha County, WI

[Pollster]

I have a busy day tomorrow so I might have to miss the Atlas meltdown that comes with this one.

https://twitter.com/monmouthpoll/status/1301274056747036675?s=21

[Pollster]

Last month, we had an extremely good poll for Biden from Monmouth in PA followed by something more tepid from them in GA and IA.

There’s no reason to expect the reverse to occur with NC tomorrow but I would like to see the amusing reaction from the forum.

To the more knowledgable polling gurus: is it actually a bad thing for a polling firm to be releasing seemingly contradictory polling? This would seem to be a sign of normal sampling error but “BOTH SIDES” (sorry, but for this it seems pretty accurate) seem to use it as an excuse to brush off polls and then ignore it when it’s favourable.

It's not a bad thing at all.  Polls *will* vary around the actual result, and an occasional one will be a wild outlier.  This is an inescapable occurrence in sampling.  An honest pollster releases whatever results they get, unless they find that something was flawed in their methodology (I've seen that happen at least once).  This demonstrates that they're not herding.  

I also recall a poll in the last cycle from one of the high quality pollsters (Monmouth, I think) that was an obvious outlier.  They released it anyway, but included a caveat to the effect of "we think this is an outlier, but it's what we got, so take it for what it's worth."

With Monmouth, the issue (for Atlas at least) is that their sample sizes are usually small, making them prone to large but ultimately statistically insignificant swings due to high margins of error that are even higher for subsamples.

Polls traditionally have a 95% confidence level (some are higher, some lower, this is usually noted in the methodology section) meaning that out of 20 methodologically equivalent polls that one pollster produces, one is likely to be an outlier. This isn’t distributed evenly, so it’s certainly possible that a pollster can put two outliers out in a row, and then 38 non-outliers, or some other combination of the ratio.

With regards to one pollster putting out contradictory numbers, there’s a lot going on there. The only true way to produce separate polls of PA and NC that would be indicative of the same national environment would be to use the exact same methodology for both of them down to a T (sampling frames, field dates, vendors, etc), have representative samples for both, and use the same post-stratification weights. And both would still be vulnerable to MOE and other systemic polling error. Many pollsters go to great lengths to make these accommodations, but public pollsters often have to cut corners because of costs.

[Pollster]

It will be hilarious when the pundit roundtables talk about how unexpected and "important" Biden's upper-single/low-double digit win in Minnesota is when it is called by the networks.

[Pollster]

The last set of Quinnipiac polls for this set of states was registered voters, the ones coming today appear to be likely voter polls.

Please keep this in mind when analyzing the trends, I shout into the Atlas void.

[Pollster]

Not sure if it's been posted here yet, but it looks like election twitter successfully crowdfunded a SurveyUSA poll of Indiana.

Results expected around October 14.

[Pollster]

https://twitter.com/Johnrigsby128/status/1318185519826743298

Don't pollsters identify themselves when they call you? I.e., "this is Quinnipiac, will you take a survey?"

IME (having been polled many times) the callers usually identify their organization, but not always.  Sometimes the introduction is like "My name is John and I'm conducting a survey on the upcoming election.  Would you be willing to answer a few questions?"

It depends on state law, in some states polls are considered campaign activity and the funder must be identified at some point (usually at the very end, we'll through something on like "paid for by XX for Congress." The actual polling company doesn't usually identify themselves unless it's a public/educational organization like Gallup or Quinnipiac.

[Pollster]

OurProgress teasing a Mississippi poll released October 25 and a Montana poll released October 26.