2020 Poll Hype Thread

[n1240]

https://twitter.com/MonmouthPoll/status/1232700397703311361

[n1240]

https://twitter.com/mulawpoll/status/1233066599067856898?s=21

[n1240]

https://twitter.com/mikepnoble/status/1268751356296179714?s=21

[n1240]

I've been analyzing polls and I've noticed a few things:

Biden's lead isn't as big as it is advertised, and in some states where Biden is polling above Trump, Trump is winning. Many polls do not weight Party ID correctly, which has to be done so there is correct sampling. Other polls oversample an urban area (more Democrat) and some do oversample rural areas. For example, Detroit was 16% of Michigan's 2016 votes, yet a poll sampled 49% from Detroit without proper weighting.

Another point I should bring up - don't look at national polls, because that is the popular vote, which Trump lost in 2016. It is not a good indication of who is winning.

Finally, I'd like to say that polls sampling Likely Voters are better indications of the 2020 race, because that includes those who are perhaps younger and are voting for the first time. Additionally, not all registered voters will vote.

Assuming you're referring to this poll that has "49% from Detroit". There is a difference between media markets and cities, the Detroit media market made up 49.7% of the Michigan electorate in 2016, so it isn't unreasonable to have a poll that has the Detroit media market at 49% (pollsters wouldn't ever weigh by area anyways, it's just demographic information they include). Good pollsters won't weigh by party ID either given how fluid that metric is (literally ask anyone from 2012 how criticizing pollsters for "oversampling" Democrats worked).

[n1240]

https://twitter.com/gravismarketing/status/1274351164490907648?s=21

[n1240]

https://twitter.com/QuinnipiacPoll/status/1283460616535998464

[n1240]

https://twitter.com/GravisMarketing/status/1283834847883124737

Re: Fox polls - possible they might have some tonight or a week from now since they usually release on Thursdays if my memory serves correct

[n1240]

https://twitter.com/electionsmith/status/1291473686525026311

[n1240]

https://twitter.com/MonmouthPoll/status/1292930228617191425

[n1240]

https://twitter.com/nate_cohn/status/1306764065394634755?s=21

Also oversample in ME-02 + house race.

[n1240]

It's probably something like

PA +5
OH +1
MOE +/- 5.5

Margin of Error acts in both directions, the margin of error indicates the error on estimates for each individual candidate, so a margin of error of 5.5 means a candidate with 50% support has a 95% confidence level that the individual candidate's support is between 44.5-55.5%, thus anything within 11% in margin between candidates can be considered within the margin of error. Assuming the pollster isn't incorrectly interpreting the margin of error to be applied in one direction (which would be a grave error on their part), it should imply that Biden's lead is high single digits/low double digits in WI/MI and anywhere between low and mid single digits in PA and OH. Their last series of polling had MOE of about 4% for reference. Pollster made a clarification that MI is outside MOE, WI just outside MOE, and PA is a larger lead than OH.

[n1240]

With 100 voters sample for every district these polls will be a) prone to wild swings, and b) almost useless.

Their last set of IA CD polls actually had about n=400 each because they had an additional set of 1200 voters who were surveyed online and only given a house preference question.

[n1240]

Monmouth’s FL poll was released at 1 pm EST, so I guess this one’s going to be released at the same time.

It's usually semi-random but I don't think they ever post beyond 1 pm - they always release at 11 am, 12 pm, or 1pm EST/EDT on the dot as far as I know.