Looking at the Civiqs poll and Gravis, Ive started to notice terrible problems in this Gravis's methodology(what else is new).
Looking at how things are broken down by race
Gravis:D/R
Whites-40/54
AA- 66/33
Latinos-31/59
Asians-63/36
Civiqs: D/R
Whites-29/65
AA-78/14
Latinos-70/25
Asians not shown
The fact that Gravis shows Latinos as lean R, African Americans as a 2/3s group, and Whites as almost even makes me lose hope that this poll really is good.
Civiqs however has pretty believable demographic numbers, though I wish that they showed Asians.
Good analysis of Gravis poll. The poll is sort of odd.
Civiqs seems an outlier. But maybe it is on to a revolution.
Before I judge any more polls I want to see how Ohio 12 plays out in August.
The Marist and YoUGove polls in Florida at about the same date with such different results is disturbing.
Are there realignments going on that are hard to poll?
This poll was a disappointment to say the least. The AZ poll had great crosstabs and made sense, it was good quality. This poll is anything but, and reminds me of that Peterson poll that the candidate is still pedaling.
Ohio 12 will be interesting to watch, and it will be interesting to see if the burbs here will swing in any way. Its usually good, however, to wait until around September to start looking at every poll and GB shift. Thats the time when most undecideds make up their mind, and can shift elections into waves.
The Marist and YouGov differences are probably due to methodology. You can see it here with the Civiqs poll and Gravis poll. Both had different results per demographic and polled different amounts of them.
Polling's usual weakness is new voters and dropoff voters. Polling relies on taking the previous electorate and using it to predict elections. The problem is, however, that new voters keep getting added in, and old voters/apathetic voters dont vote. This can lead to polls not figuring out about waves until they crash. That was the case in 2010 and 2014. Will they miss a demographic swing? No. Will they miss what percentage of the electorate is which demographic? Yes. But as time has gone on, we have gotten better at polling, so mistakes are becoming smaller and smaller.
