Thursday, October 30, 2008

Tracking the Tracking Polls

Zogby, Rasmussen Reports, Quinnipiac, Gallup--we've heard the names of the polls and what they are predicting, but what about the polls themselves? To learn about what kind of samples they use, when they publish findings, and their strengths and weaknesses, see FiveThirtyEight.com for a Tracking Poll Primer written by Nate Silver (an "all-star in the world of baseball stats). His site is named for the number of electors in the electoral college.

Here are some more resources for you. Come on, I know you're fixated on this election, just as we all are. Might as well pass the time productively...

Pollster.com

Rasmussen Reports, Presidential Election Polls

RealClearPolitics

USA Today's Presidential Poll Tracker

After you've spent half your day looking at statistics, take a look at this article: General Election Polls: A History of Inaccuracy; The Sad History of General Election Polls, and How They Have Repeatedly Failed to Predict the Outcomes of Presidential Elections. It was written in February, 2008 by Nithin Coca, a free-lance writer for Associated Content. Note that this website is an "open content network," allowing anyone to submit content, so you might check Nithin's stats.

Wait, you probably just want to cut to the chase, so I'll quote the entire article right here for you.

I'm sure you've noticed all the political campaign touting general election polls. Unfortunately, these polls have a terrible history of actually predicting who will in the fall. So what is Barack Obama leads everyone in Zogby, woop-dee-do! Does John Edwards leads in Rassmussen, oh my lord! Clinton leads in ARG? Yowsie!

I'm going to explore how the polls have failed repeatedly, and show you the real margin of error. So next time you see a poll, read it with caution!

Here are some of the worst disasters of General Election polling from the last 24 years of Presidential elections.

After this January's debacle in New Hampshire, can we just argue on the issues and the REASONS why to support a candidate, and ignore faulty polls?

1976Late July - Gallup
Jimmy Carter 62%
Gerald Ford 30%

Final Results
Carter 50.1%
Ford 48.0%

Average MOE - 14.95% This sort of shift would make it a blowout for either side of the aisle.

1980 (this one's for those of you who say - "polls shift over time") Nov 1980, Gallup Pre-Election Poll
Carter 44%
Reagan 41%

Final Results
Reagan 50.7%
Carter 41.0%
Average MOE - 5.85% = the margin of error in every GE poll this year. This really embarrassed the pollsters, so of course, they went ahead and did it again.

1988
5/17 - NYT/CBS
Michael Dukakis 49%
George Bush 39%

Final Results
Bush 53.4%
Dukakis 45.6%

Average MOE - 7.9% A shift like what occurred in 1988 would make any Democrat the winner or the loser by a healthy margin.

1992
June 1992 Time/CNN
Ross Perot 37%
George Bush 24%
Bill Clinton 24%

Final Results
Clinton 43.0%
Bush 37.4%
Perot 18.9%

Average MOE - 20.1%. Imagine if Bloomberg's runs, I foresee similar dynamics.

2000
Sept 2000 Newsweek
Al Gore 49%
George W. Bush 39%

Final Results
Bush 47.9%
Gore 48.4%

Average MOE - 4.8% So all the undecided went for Bush, eh? Polls are worthless in close races. Hmmm, sounds familiar, doesn't it?In conclusion, the only poll that matters is the one on election day.

2 comments:

Margie's Musings said...

Scary, isn't it? That's why I just discount the polls. There's no way to know until the fat lady sings.

Erikka said...

funny voting story - i ordered my absentee ballot from Candia and they mailed it the 17th of November. Have I received it? No. They have now mailed me another one (as of yesterday) that needs to be back to them by Tuesday...

Conspiracy theory!?!?!