Which pollster is the most accurate?

[ukchris82]

I'm into American politics, but have nothing on you guys/gals.
I look at the polls and I'm unsure which ones are worth taking note of and which aren't.

Which pollster is the most accurate?
Which are partisan?

Cheers

Chris

[I spent the winter writing songs about getting better]

Which pollster is the most accurate?

Overall Mason-Dixon. Rasmussen has been the best the last few cycles.

Which are partisan?

Anything labeled an internal.

[Eraserhead]

Beware of just about anything from Zogby and ARG.

[AndrewTX]

zogby is the most accurate, hands down.

 

[Verily]

zogby is the most accurate, hands down.

LOL



Zogby is the worst. They're not partisan, just bad.
ARG is also bad. They tend to weight their figures in favor of who they think is leading (so, for example, they had McCain leading the Republican primary in Iowa long after everyone else, and the same for Giuliani when Romney surged in Iowa).

Survey USA and Research 2000 are middling quality, but both commonly perform surveys for specific organizations, and you have to watch to see if there's a partisan or ideological slant to the organization hiring them.

Mason-Dixon and Rasmussen are the best, though Rasmussen has had a documented history of slightly favoring the Republicans long in advance of an election (though not when an election approaches). Rasmussen also uses a different approval rating formula than everyone else.

Other surveys tend to only cover a few states. Generally, universities are terrible, unless they are Quinnipiac College (which is middling) or the University of New Hampshire (which is great in New Hampshire). The Field Poll in California is gospel, but most local polls are unreliable.

[Mr. Morden]

Zogby is the worst. They're not partisan, just bad.
ARG is also bad. They tend to weight their figures in favor of who they think is leading (so, for example, they had McCain leading the Republican primary in Iowa long after everyone else, and the same for Giuliani when Romney surged in Iowa).

Is Zogby really worse than ARG?  (I'm only talking about Zogby's telephone polls here.  The internet polls are definitely worthless.)  I mean, at least for the 2008 primary cycle, ARG has been terribly out of synch with every other pollster in most of their state polls.  Far more so than Zogby.  (That doesn't necessarily prove that ARG is worse, but it's not a good sign.)

[Eraserhead]

zogby is the most accurate, hands down.

LOL



Zogby is the worst. They're not partisan, just bad.
ARG is also bad. They tend to weight their figures in favor of who they think is leading (so, for example, they had McCain leading the Republican primary in Iowa long after everyone else, and the same for Giuliani when Romney surged in Iowa).

Survey USA and Research 2000 are middling quality, but both commonly perform surveys for specific organizations, and you have to watch to see if there's a partisan or ideological slant to the organization hiring them.

Mason-Dixon and Rasmussen are the best, though Rasmussen has had a documented history of slightly favoring the Republicans long in advance of an election (though not when an election approaches). Rasmussen also uses a different approval rating formula than everyone else.

Other surveys tend to only cover a few states. Generally, universities are terrible, unless they are Quinnipiac College (which is middling) or the University of New Hampshire (which is great in New Hampshire). The Field Poll in California is gospel, but most local polls are unreliable.

I'm guessing Andrew was joking.

[Verily]

zogby is the most accurate, hands down.

LOL



Zogby is the worst. They're not partisan, just bad.
ARG is also bad. They tend to weight their figures in favor of who they think is leading (so, for example, they had McCain leading the Republican primary in Iowa long after everyone else, and the same for Giuliani when Romney surged in Iowa).

Survey USA and Research 2000 are middling quality, but both commonly perform surveys for specific organizations, and you have to watch to see if there's a partisan or ideological slant to the organization hiring them.

Mason-Dixon and Rasmussen are the best, though Rasmussen has had a documented history of slightly favoring the Republicans long in advance of an election (though not when an election approaches). Rasmussen also uses a different approval rating formula than everyone else.

Other surveys tend to only cover a few states. Generally, universities are terrible, unless they are Quinnipiac College (which is middling) or the University of New Hampshire (which is great in New Hampshire). The Field Poll in California is gospel, but most local polls are unreliable.

I'm guessing Andrew was joking.

I guessing I knew that, but I'm not sure.

[RJ]

Mason-Dixon and Rasmussen are the best, though Rasmussen has had a documented history of slightly favoring the Republicans long in advance of an election (though not when an election approaches). Rasmussen also uses a different approval rating formula than everyone else.

I'm glad I'm not the only one who has noticed this.

[Gustaf]

In Zogby's defence IIRC they weren't THAT bad in 2006. Not good, but not in their own class compared to say ARG or SUSA.

[Gustaf]

Mason-Dixon's deviation:

NJ: 1.5%

PA: 4.3%

MI: 0.4%

WA: 3%

OH: 5.5%

AZ: 0.77%

MO: 1%

MD: 8%

RI: 8%

TN: 9%

VA: 0.7%

MT: 0.7%

UT: 2%

FL: 1.7%

NV: 2.5%

Average: 3.2%, wrong calls: 1 (RI)

In my opinion, not that good, but also not really disastrous. The real screw-up is really RI (+TN and MD). Beyond that they were mostly ok. In MD's defence it should be noted that these polls were half-a-week old by election day (though there are few indications of major changes in the races during that time, so it isn't much of a defence).

Zogby:

FL: 6.7%

RI: 7%

VA: 0.7%

PA: 9%

TN: 7%

OH: 4.6%

NJ: 6.5%

MT: 0.3%

MO: 1%

MD: 6%

CT: 1.8%

TX: 9%

MI: 11.5%

NY: 10%

Average: 5.8%, wrong calls: none!

Yeah, Zogby still pretty much sucks even though they were ok on most competitive races.

Rasmussen:

MO: 3%

MT: 1.3%

TN: 1.3%

VA: 0.3%

NJ: 0.5%

MN: 6%

MD: 6%

NV: 2.4%

PA: 4.3%

CT: 2.1%

WA: 8%

Average: 3.2%, wrong calls: 1 (MO, Virginia I count as a correct call)

THe problem here is that the polls at the bottom of the list are more than a week old. For some reason Rasmussen hardly polled any races really close to the election. THe ones that are really close are more spot on. So it seems Rasmussen "won" this round.

Gallup:

MO: 2%

TN: 0.2%

RI: 4%

NJ: 4.5%

MT: 8.3%

VA: 3.3%

Average: 3.7%, wrong calls: 1 (Virginia)

Also not too shabby, but this is based on a lot fewer races, so that kind of goes against them IMO.

Someone with more time should of course make one with statistical variance, etc. And include the other pollsters.

[Eraserhead]

MD kind of blew TN last year.