FL/OH/PA Quinnipiac: Clinton +1 in FL, +4 in OH, +10 in PA

[Wiz in Wis]

FL: Clinton 46 – Trump 45
OH: Clinton 49 – Trump 45
PA: Clinton 52 – Trump 42

https://twitter.com/samsteinhp/status/763081941134565376

PA is clearly gone.  FL seems oddly competitive, but then again, this is Q-poll, so probably still a bit Trump friendly.

[Wiz in Wis]

For Context
FL: Clinton 46 – Trump 45 - Was Clinton 39 – Trump 42 (4 pt swing)
OH: Clinton 49 – Trump 45 - Was Clinton 41 – Trump 41 (4 pt swing)
PA: Clinton 52 – Trump 42 - Was Clinton 41 - Trump 43 (12 pt swing)

[Wiz in Wis]

https://twitter.com/aedwardslevy/status/763085018163970052

this is Quinnipiac's first survey using a likely voter model in these states. Past were all registered voters

I wonder if that helps or hurts her... this year, LV screen has been unusually pro-Dem.

[Wiz in Wis]

Can we see some crosstabs, specifically Florida

https://www.qu.edu/news-and-events/quinnipiac-university-poll/2016-presidential-swing-state-polls/release-detail?ReleaseID=2371

Ironically, the race tabs in FL look right, but they are over-sampling whites in both OH and PA by a few points compared to 2012. Of course, in all 3 states, they still estimate the electorate to be AT LEAST as white as 2012... none of these samples has oversampled minorities.

[Wiz in Wis]

Can we see some crosstabs, specifically Florida

https://www.qu.edu/news-and-events/quinnipiac-university-poll/2016-presidential-swing-state-polls/release-detail?ReleaseID=2371

Ironically, the race tabs in FL look right, but they are over-sampling whites in both OH and PA by a few points compared to 2012. Of course, in all 3 states, they still estimate the electorate to be AT LEAST as white as 2012... none of these samples has oversampled minorities.

But just stop it...

This survey uses statistical weighting procedures to account for deviations in the survey
sample from known population characteristics, which helps correct for differential survey
participation and random variation in samples. The overall adult sample is weighted to
recent Census or American Community Survey data using a sample balancing procedure to
match the demographic makeup of the population by county, gender, age, education and
race. Margins of sampling error for this survey are not adjusted for design effect.

Stop what? I am just pointing out that these polls are NOT oversampling minorities.