FiveThirtyEight releases pollster ratings

[JRP1994]

http://fivethirtyeight.com/interactives/pollster-ratings/

[eric82oslo]

They "forgot" to include Gravis Marketing.

I'm surprised at how well Gallup does considering how horribly wrong they were in their assumptions ahead of the 2012 election. They were like constantly thinking that Romney would win by 7-8%, while he in reality lost by 4%.

[King]

Gallup openly admitted what was wrong with their models and have corrected it as opposing to doubling down a la Rasmussen.

[Vosem]

They "forgot" to include Gravis Marketing.

Uh, no they didn't...

Gravis Marketing
◐
20
3.9
-0.0
+0.6
+0.6
C
R+1.4

[eric82oslo]

They "forgot" to include Gravis Marketing.

Uh, no they didn't...

Gravis Marketing

They're not included in the graphic part of the article at least. I guess they are in the linked document/article though.

[eric82oslo]

It's interesting to see that all of the five most common pollsters have a Republican bias, PPP included. The five are: SurveyUSA, Rasmussen, Mason-Dixon, Zogby and PPP. And out of the 13 most common pollsters, only one pollster had a Democratic bias; Research 2000, which has been banned by 538. Out of the 16 most common pollsters - all with between 80 and 700+ individual polls analyzed, only two had on average a Democratic bias - the other one being InsiderAdvantage with a slight 0.3% Democratic bias. Between these 16, we had several pollsters with a strong Republican bias:

TCJ Research at R +4.5%
Rasmussen Reports at R +2.3%
Strategic Vision at R +1.9%
Harris Interactive at R +1.7%
Marist College at R +1.5%
Zogby Interactive at R +1.4%
YouGov at R +1.2%
Mason-Dixon at R +1.1%
Quinnipiac at R +0.9%
Gallup at R +0.8%

Also:

We Ask America at R +1.5%
Gravis Marketing at R +1.4%

[eric82oslo]

538's ranking of the very best and brightest pollsters in America (only including pollsters who have fielded at least 20 polls):

1. Selzer & Company: A+ : 0.0% bias/no bias
2. Field Research Corporation/Field Poll: A+ : 1.1% R bias
3. SurveyUSA: A : 0.2% R bias
4. Grove Insight: A : 0.5% D bias
5. Monmouth University: A- : 0.1% R bias
6. Ipsos, telephone: A- : 0.4% R bias
7. NBC News/Wall Street Journal: A- : 0.5% D bias
8. Muhlenberg College: A- : 0.5% R bias
9. CNN/Opinion Research Corp.: A- : 0.6% R bias
10. ABC News/Washington Post: A- : 0.6% D bias

11. Siena College: A- : 0.8% R bias
12. University of Cincinatti/Ohio Poll: A- : 1.5% R bias
13. Los Angeles Times: B+ : 0.6% R bias
14. Research & Polling Inc.: B+ : 0.6% R bias
15. Quinnipiac University: B+ : 0.9% R bias
16. Mason-Dixon Polling & Research: B+ : 1.1% R bias
17. Marist College: B+ : 1.5% R bias
18. Market Shares Corporation: B+ : 1.7% R bias
19. Princeton Survey Research Associates International: B : 0.2% D bias
20. RKM Research & Communications: B : 0.2% R bias

21. EPIC-MRA: B : 0.3% R bias
22. CBS News/New York Times: B : 0.7% D bias
23. Suffolk University: B : 1.1% D bias


Meanwhile, Gallup is only ranked with a C+ mark, while Rasmussen and Gravis both do even worse, at only C.

[Miles]

Why did they ban Pharos? Other than showing Bob Kerrey within the MoE, I remember them being pretty good in 2012.

[eric82oslo]

I wonder how you can get an A+ when you have a 1.1% R bias. That's at least as strong and probably an even stronger bias than the average pollster. It's only slightly less of a bias than Gravis even.

[eric82oslo]

When it comes to other prominent pollsters, 538 have ranked them:

Pew Research Center (B) at about 35 or below
PPP (B-) at about 62 or below
Fox News (C+) at about 91st or below
Gallup (C+) at about 92nd or below
YouGov (C+) at about 93rd or below

This (C+ field) is also where you'll find Landmark, Google, Rutgers, Harris Interactive and Magellan Strategies.

Worst non-banned pollster according to 538: Zogby Interactive/JZ Analytics. The only non-banned pollster with an F mark. Second worst: Millersville University with a D- mark.

Rasmussen Reports has the 5th strongest bias of all 300+ pollsters. POS (Public Opinion Strategies) has the 6th strongest bias. Gravis is at place 21. That is, amongst the less than 10% of the very worst pollsters (when it comes to bias at least).

Roanoke College, surprisingly, has a 0.0% bias, yet only gets a C- rating.

[Obama-Biden Democrat]

They "forgot" to include Gravis Marketing.

I'm surprised at how well Gallup does considering how horribly wrong they were in their assumptions ahead of the 2012 election. They were like constantly thinking that Romney would win by 7-8%, while he in reality lost by 4%.

They blew 2008 as well , they had Obama up by 12-15 points in 2008, and 2010 they had the GOP up 15.

[7,052,770]

I wonder how you can get an A+ when you have a 1.1% R bias. That's at least as strong and probably an even stronger bias than the average pollster. It's only slightly less of a bias than Gravis even.

I assume if the pollster is consistently, always dead-on 1.1 R-biased, then you can get a very accurate reading from the poll by making the adjustment.

That's a lot more reliable than one that averages out to 0, but has a high standard deviation.

[I spent the winter writing songs about getting better]

The fact that PPP actually has a slight R bias is something I'd love to rub in the face of certain hacks except they all disappeared on November 7, 2012.

[Free Bird]

I'm tasting the tears of the PPPites right now.

[○∙◄☻¥tπ[╪AV┼cVê└]

The fact that PPP actually has a slight R bias is something I'd love to rub in the face of certain hacks except they all disappeared on November 7, 2012.

That's because the November 6, 2012 polls oversampled minorities. You need to unskew them.

[IceSpear]

Interesting that almost all pollsters had at least a slight Republican bias. In very close races, that could be decisive.

[SteveRogers]

I'm tasting the tears of the PPPites right now.

Um, how so?

[7,052,770]

I'm tasting the tears of the PPPites right now.

What? PPP got a pretty good rating from 538, but not nearly enough for any tears of joy.

If anything, it's probably the PPP-haters crying sad tears, but 538 demonstrated how their two pet arguments ("PPP is a bad pollster," "PPP has a Democratic bias") are wrong.

[Tender Branson]

As I have said before, trust SurveyUSA more and ignore the joke polls.

[eric82oslo]

As I have said before, trust SurveyUSA more and ignore the joke polls.

I certainly trust them now that Crist is well ahead, once again.