Why several polls are wrong - Gallup, CBS, etc...
Posted: Thu Oct 21, 2004 3:41 pm
http://www.grassrootsbuzz.com/archives/000291.html
Grassroots Buzz
September 18, 2004
Why The Polls Are Wrong
NOTE: This post is long but the info is extremely important. Biased polls can disillusion a candidate’s supporters and we need to make sure everyone knows that Bush is going to lose this race. Pass it on.
I’m not a polling expert by any measure but some simple analyst of the numbers shows why the recent Gallup and CBS/NYTimes polls are way off the mark. It all comes back to their political sample; the number of Democrats/Republican/Independents the pollster assumes will come out to vote. To me, the best estimates of turnout are based on past Presidential elections.
1996: 39% Democrats, 34% Republicans, and 27% Independents.
2000: 39% Democrats, 35% Republicans and 26% Independents.
Both those elections had very different dynamics with very different results, but turnout was about the same. Gallup's latest poll gives Bush a 13 point lead among likely voters, winning 55% to 42%. That is a blow-out. It is also completely different from some other polls that have come out this week. Pew Research and Democracy Corp (partisan) called it a 1% point difference and Zogby says 2% difference.
Steve Soto at The Left Coaster asked Gallup for their polling sample and actually received an answer. For some unknown reason, Gallup is weighing the election to reflect an enormous Republican turn-out.
Total Sample: 767
GOP: 305 (40%)
Dem: 253 (33%)
Indy: 208 (28%)
I have no idea why Gallup would estimate more Republicans voting than has happened in the past 20 years (could be longer but I don’t have the numbers). Actually I do have an idea, James Clifton, Gallup’s CEO, has donated thousands to Republican canidates. I don’t think that automatically makes his polls suspect but until there is a better explanation what else can one think.
This isn’t the first time Gallup has shown an abnormal Bush lead. In mid August (8/11), Gallup put Bush ahead by 3 %, while Zogby, Time and Democracy Corp gave Kerry a 7% lead and Fox predicted Kerry was ahead by 5%.
The same sampling problems are reflected in the CBS/NY Times poll which shows an 8% lead for Bush, 50% to 42%. In their sample, they have a huge suppression of Democratic voters and an gigantic leap in independents.
Total Sample: 1287
GOP: 426 (34%)
Dem: 399 (31%)
Indy: 462 (36%)
Why god, why? Donkey Rising reweighed the numbers to reflect the 2000 turnout and found the race to be tied.
Grassroots Buzz
September 18, 2004
Why The Polls Are Wrong
NOTE: This post is long but the info is extremely important. Biased polls can disillusion a candidate’s supporters and we need to make sure everyone knows that Bush is going to lose this race. Pass it on.
I’m not a polling expert by any measure but some simple analyst of the numbers shows why the recent Gallup and CBS/NYTimes polls are way off the mark. It all comes back to their political sample; the number of Democrats/Republican/Independents the pollster assumes will come out to vote. To me, the best estimates of turnout are based on past Presidential elections.
1996: 39% Democrats, 34% Republicans, and 27% Independents.
2000: 39% Democrats, 35% Republicans and 26% Independents.
Both those elections had very different dynamics with very different results, but turnout was about the same. Gallup's latest poll gives Bush a 13 point lead among likely voters, winning 55% to 42%. That is a blow-out. It is also completely different from some other polls that have come out this week. Pew Research and Democracy Corp (partisan) called it a 1% point difference and Zogby says 2% difference.
Steve Soto at The Left Coaster asked Gallup for their polling sample and actually received an answer. For some unknown reason, Gallup is weighing the election to reflect an enormous Republican turn-out.
Total Sample: 767
GOP: 305 (40%)
Dem: 253 (33%)
Indy: 208 (28%)
I have no idea why Gallup would estimate more Republicans voting than has happened in the past 20 years (could be longer but I don’t have the numbers). Actually I do have an idea, James Clifton, Gallup’s CEO, has donated thousands to Republican canidates. I don’t think that automatically makes his polls suspect but until there is a better explanation what else can one think.
This isn’t the first time Gallup has shown an abnormal Bush lead. In mid August (8/11), Gallup put Bush ahead by 3 %, while Zogby, Time and Democracy Corp gave Kerry a 7% lead and Fox predicted Kerry was ahead by 5%.
The same sampling problems are reflected in the CBS/NY Times poll which shows an 8% lead for Bush, 50% to 42%. In their sample, they have a huge suppression of Democratic voters and an gigantic leap in independents.
Total Sample: 1287
GOP: 426 (34%)
Dem: 399 (31%)
Indy: 462 (36%)
Why god, why? Donkey Rising reweighed the numbers to reflect the 2000 turnout and found the race to be tied.