Poll analyst's advice: Don't believe internal polls from either party

[Valienteman]

"You are a fool" to do so.

http://twitter.com/LarrySabato/status/27917195557

[HopeHoops]

1. The first part of "analyst" is "anal". Don't even get me STARTED on the "poll" part.

[Valienteman]

4. Larry Sabato's history

http://en.wikipedia.org/wiki/Larry_Sabato

"Prior to the 2002 midterm elections, where the Republican Party saw gains in both branches of Congress, Sabato's Crystal Ball website accurately predicted the outcome in 433 of the 435 contests for the House of Representatives and 32 of 34 Senate races.<8>

"In 2004, which saw Republicans retain the White House and gain seats in the House and Senate, Crystal Ball correctly predicted the outcome of 525 of the 530 political races (99% accuracy), missing only one House race, one Senate race, one governor's race and two states in the Electoral College.<9>

In August 2006, Crystal Ball predicted that the Democrats would gain 29 seats in the House of Representatives and 6 seats in the Senate, providing them with a majority in both chambers. Sabato's predictions proved correct: each of his 33 Senate predictions were accurate, and in the House, Democrats gained 29 seats on election night, the precise total predicted by the Crystal Ball (Democrats would go on to pick up a 30th seat in the December 12, 2006 run-off in Texas' 23rd district).<10>

In July 2008, Crystal Ball correctly projected that Barack Obama would win the presidency in a near-landslide.<11> Sabato predicted a 364-174 margin in the Electoral College, as well as the popular vote percentages.<12> The prediction was merely one point off the mark, with the actual result on November 4, 2008 being Obama 365 and McCain 173. Crystal Ball also accurately predicted 100% of all 35 Senate races, and 11 gubernatorial races correctly.<13>

In 2006 Sabato was named the most accurate prognosticator by FOX News, MSNBC, CNBC, and Pew's Project for Excellence in Journalism. In 2006, Sabato was the only national analyst to correctly predict the exact Democratic gains in Senate and House contests."

[jobycom]

7. Yeah, but the last part is "st," which of course means "Saint."

Are you really questioning the opinion of an anal, listing Saint?

[HopeHoops]

8. Not if he's stoned. And continuting on the theme, my college course "Real Analysis"....

... was listed on my report card and transcript as "REAL ANAL" - I've always said it had me bent over all semester.

[jobycom]

11. Stoned like Spicoli or stoned like Stephen?

And that's a classic report card. :)

[Kurt_and_Hunter]

2. And the reason not to do so is not because the polls are cooked...

...but because campaigns decide which polls to release.

(The polls are usually as methodologically good as any other.)

[The Northerner]

3. I concur

[emulatorloo]

5. As opposed to Gallup, whose Likely Voter model is that minorities are going to go big for the Teabag

Party,

[Radical Activist]

6. You're a fool to believe the corporate media polls.

[bluestateguy]

9. Karl Rove did this on the eve of Election 2000

He pimped secret polls to Matt Drudge showing Bush ahead in Pennsylvania and a dead heat in California.

[jobycom]

10. To paraphrase, "It's a wise person who only believes half of the polls they see...

and a brilliant person who believes the right half."

Polls are only as accurate as the formulas used and the questions asked. Campaign polls are probably more accurate than a lot of the corporate polls because campaigns want to know what to target, but campaigns aren't going to release them unless they help. Corporate polls have different agendas. Some want to elect one party or the other, some want to make outrageous predictions so the media will buy them for the shock value, some want to make the elections seem close so that the media will buy them for the cliffhanger value.

Overall, though, you can graph out all the polls and get a fair idea of what's happening. If all the polls show one candidate winning by a lot, chances are good that they are. If all the polls show a close race, chances are the race is close. If the polls show a close race with one candidate winning, and the polls you are looking out come from different sources with different agendas and different formulas, chances are they are close to right.

Polls are very useful, but very often misused.