“Not really. It is only tough if you use weighting for your models. SurveyUSA does true random samples and does not weight by party or demographics. This is why movement is picked up by their surveys and seem in my opinion to be most trusted..even when I dont like..”
***
That whole “Party ID weighting” gambit
is beloved by all the pros, but also provides
a loophole big enough to drive an opinion thru
... for any pollster so inclined to mold opinion
rather than simply report it.
Pollsters are welcome to report their Margins Of Error as what formal statistics suggests ... “+/- 3%” and the like.
The informed user would do better to apply
actual empirical results like those posted, for each separate poll.
When all eyes turn to Ohio on Election night, the only poll to pay attention to will be the University of Cincinnati poll, in my opinion.
Here’s a sampling of their recent work:
2004 Presidential race:
Final University of Cincinnati poll (10/27-10/31)
50.1% Bush
49.2% Kerry
Actual election results
50.82% Bush
48.71% Kerry
2006 U.S. Senate
Final University of Cincinnati Poll (11/01-11/05)
56% Brown
44% DeWine
Actual election results
56% Brown
44% DeWine
2008 Ohio Democratic Primary
51% Clinton
42% Obama
Actual election results
54% Clinton
44% Obama