Non-Gallup/Rasmussen tracking polls thread

[kevinatcausa]

Didn't Zogby have a big jump towards McCain 3 days ago?  If so, there may be a jump back today as that sample drops off. 

[kevinatcausa]

Even Georgia's looking a little bit less like an outlier as the night goes on.   The last returns coming in seem to be the early voting from Fulton and Gwinnett counties (metro Atlanta), and McCain's lead is already down to 7%. 

[kevinatcausa]

Based on the current national lead of 6.2% for Obama,

Pew, CNN, Ipsos/McClatchy, Hotline, FoxNews, Rasmussen, and Battleground (Lake) were ALL within 1.2% of the correct number. 

All of these except Rasmussen and Battleground are included in the New York Times polling average, which specifically excludes polling firms that use robot calling. 

The average of the 15 pollsters in the RCP average (7.3%) is only off by about 1% (The RCP average of 7.6% is slightly higher since they give the two Battleground polls 50% weight). 

The 538 projection looks to be within 0.1%, and its creator specifically said he did not include any sort of Bradley effect because he did not believe one existed. 

[kevinatcausa]

The Ipsos/McClatchy number I'm going off of is the one once they push leaners:  53-46 Obama.  This nails McCain exactly, and is within 1 on the margin (even the 8% before leaners are included only 1.8% off, the MoE is 3.6%).

You can't compare McCain's 42 in the 50-42 numbers (before leaners pushed) to his 46 in the final results (where every leaner by definition decided) -- it's apples and oranges. 

Also, I'm not sure how a margin of Obama +7 from FoxNews (following the link from the RCP polling average) is "just barely within" a 3 point error margin of the current Obama +6.2 results.  

[kevinatcausa]

The projected popular vote margin posted on this site is now 7.03%, which would put CNN, FOX, and McClatchy as the spot on pollsters if it holds up. 

Maybe its worth waiting a while before we compare polling methods...or just not worry about it in a case where almost every pollster was well within the margin of error (exceptions looking like Zogby and Gallup on the Democratic side and Battleground(Lake) on the Republican side)

[kevinatcausa]

IDB/TIPP:         -2.6
Gallup:             -4.9
R2K:                 -0.9
Zogby:              -4.1 (It's Zogby!)
Hotline:            -1.9
ABC/WP:          -3.9
Ras:                 -0.4
Ipsos:              -4.9

Every one of these polls under projected McCain, though many were in the MOE.  Obama was undercounted in some, but not all.

Okay, now I'm about ready to start banging my head against the wall too.  The only number that's worth anything whatsoever from a poll that doesn't either push all the leaners or allocate all the undecideds is the margin.  Under no circumstances should you be comparing the absolute percentages of any single candidate in a poll which has undecideds with those from an actual election which doesn't.   This isn't a Bradley effect thing, this isn't a methodology thing...this is the obvious fact that even the best poll which only allocates 94% of the voters will undercount McCain. 

The way to "fix" this undercount isn't to give up and claim that other pollsters are better.  It's to stop looking at the McCain totals and focus on the margin instead!

[kevinatcausa]

One thing to watch out for:  The CNN numbers you gave add up to 100%, so I assume you're completely ignoring third parties.  The polls don't. 

My best estimate (I can't find a total anywhere right now) is that right now McCain's at 46.2-46.3% of the total vote, with 2% of the votes left to be counted (mostly from Oregon, Washington, and California).  Together, Obama and McCain make up about 98.5% of the vote. 


Some pollsters push leaners and/or allocate undecideds until they reach pretty close to that value.  Examples include the Battlegrounds (which both overestimated McCain and underestimated Obama to various extents), Gallup (which pulled a Zogby), and Rasmussen   For those polls, taking the absolute percentages is almost the same as just looking at the margin of error, and if they're off by any amount towards either candidate, it'll lead to a positive number.

Other polls (e.g. Ipsos before leaners) only allocate 92% of the vote.   All other things being equal, you'd expect an undercount of 3%+ on each candidate.  For Ipsos, it was actually 4.2, because the leaners were more heavily McCain than the general sample.  However, we knew that was going to happen in advance, because Ipsos also included results with leaners!  Given a choice, we take the results with leaners, (53-46), which also are within tenths of a percent for each candidate.   

The point is, despite that horrific looking -4.9 in your table, Ipsos was one of the best pollsters this election.  Looking at the absolute difference instead of the margin disguises this fact.   The same holds true for other pollsters on your list.  IBD/TIPP only assigned 96% of the voters, and that alone accounts for half of the -2.6 margin (your ignoring third parties accounts for another 0.75 or so).   

The two polls that have big numbers in your table that assigned leaners were Gallup and Zogby.  But those were outliers this year, as you could have seen just by looking at the margins.  The absolute differences told you nothing new. 

[kevinatcausa]

As I mentioned above, Hotline, along with most of the firms the were within 1 point of the correct margin (Rasmussen being an exception), does not use robot polling.