National Tracking Poll Thread

[greenforest32]

Those articles are not comparing the same thing. The 2012 link on the house effect (individual pollster difference from the polling average) is not the same thing as comparing a pollster's results to the election results (the 2010 link).

Obviously we can't compare the 2012 polls to the results yet as the election isn't over but the polling average is not the same thing as the election results.

https://uselectionatlas.org/FORUM/index.php?topic=155124.0

Ras is a well known troll. You don't have to look past his Senate polls to see that. Just look at Florida: http://en.wikipedia.org/wiki/United_States_Senate_election_in_Florida,_2012#Polling_2

August 15 - Nelson +7
July 9 - Mack +9
April 25 - Nelson +11
March 13 - Mack +7
February 13 - Tie
November 17 (2011) - Mack +4

[greenforest32]

So, that link you provided to "prove" that Rasmussen is a "known troll" would also indicate that Quinnipiac is a troll (a top rated pollster by Silver's standard), Mason Dixon (another good poll), and PPP (which I'm sure, you just think is great).

Neither of them has wild swings like Rasmussen showing 15+ point swing in one month (Mack +7 in March to Nelson +11 in April). The 2010 link above comparing pollsters against the election results should be proof enough no? Don't know why he removed this chart from the article though.

[greenforest32]

The point is that the measure of accuracy should be the poll against the election results, not the poll against the polling average as the average can be wrong.

Ras is clearly the odd one out in the Florida polling. They've shown two different 15+ point swings in one month while Quinnipiac had smaller swings (one over two months). Their past results show they are not credible compared to the other major pollsters. They're in their own league:

You must be logged in to read this quote.

Edit: missed the second Quinnipiac poll

[greenforest32]

Nobody is saying the 2012 differences will be the same but the point is that the last objective measure we have of the polling results is from the last election: 2010.

I don't really care one way or another but I just get annoyed at the implication that Rasmussen's inaccuracy was the same as the other pollsters or never existed or is now gone.

[greenforest32]

The point is that the measure of accuracy should be the poll against the election results, not the poll against the polling average as the average can be wrong.

If there is a "house bias," which you claimed, then that should not effect the change in the poll, only the final number.

You must be logged in to read this quote.

Here is the link you posted:  http://en.wikipedia.org/wiki/United_States_Senate_election_in_Florida,_2012#Polling_2

It shows Nelson:

Rasmussen Reports:    July 9, 2012        37%    
Quinnipiac:    June 19-25, 2012:             41%    

Then it shows Nelson:

Quinnipiac    August 15-21, 2012                    50%
Rasmussen Reports    August 15, 2012     47%    

The polls show a gap of 4 points at most.  Rasmussen shows a gain of 10 points; Quinnipiac  shows a gain of 9 points.

I wasn't talking about Quinnipiac v Rasmussen, I claimed Rasmussen had two wild one month swings in an otherwise stable race:

March 13, 2012 - Mack +7
April 25, 2012 - Nelson +11

and

July 9, 2012 - Mack +9
August 15, 2012 - Nelson +7

I could see swings like that given Todd Akin comments but nothing like that has happened in Florida. They could be outliers but I think it's more narrative setting polls given Rasmussen's track record.

There are legitimate reasons to doubt their credibility. It's nothing new.

[greenforest32]

They've been objectively better so I don't doubt them to the same degree as Rasmussen.

What I like about them is that they release all the cross-tabs and poll questions aside from the main race like state legislature ballots, approval ratings for other officials/issues, etc.

False equivalency is always nice.

[greenforest32]

Yes, and you are making a false equivalency by comparing a "house bias," i.e. how much the poll varies from the result, with swings internal to multiple polls.  As was demonstrated, both Quinnipiac and Rasmussen showed about the same shift to Nelson, roughly over the same period.

Who was conflating them? They can do both. Like they push out polls to the right of more accurate pollsters and then correct when new polls come in, that's what I mean by polls setting a narrative.

And since when do we compare shifts solely along one candidate? The numbers could change for the other candidate thus leading to a wider gap between them and for that gap Quinnipiac is lower (+1 Nelson to +9 Nelson) compared to +9 Mack to +7 Nelson for Rasmussen. The same thing happened in November-April.