Gallup Poll now officially laughingstock
Gallup's most recent poll has finally proved that they cannot ever be taken seriously. Ever. After publishing a poll with a Republican bias of +7%, they now have a poll with R+12% bias which gives, of course, Bush Retains Lead, Up by Eight Points. Just to reiterate, the polls in question have internal respondents which profess Party Identification (Party ID) with a Republican bias of 7% and 12%. This means there are 7 and 12 percent more Republicans than Democrats in the survey. The problem with this is that these two polls' internals do not reflect real life.
Lets look at some internals of the R+7% (13-15 Sept) and the R+12% (24-26 Sept):
Likely Voter Sample Party IDs – Poll of September 13-15 (R7% bias)
Reflected Bush Winning by 55%-42%
Total Sample: 767
GOP: 305 (40%)
Dem: 253 (33%)
Ind: 208 (28%)
Likely Voter Sample Party IDs – Poll of September 24-26 (R12% bias)
Reflected Bush Winning by 52%-44%
Total Sample: 758
GOP: 328 (43%)
Dem: 236 (31%)
Ind: 189 (25%)
I ask anyone to bring data forward that Republicans have had a 12% bias in the last three decades, much less the last four years. Look, hard, because you won't find it. Luckily, here is the Historical Party ID from 1987-2004 (pdf):
| Rep. | Dem. | Ind/Oth | Dem Adv. +/- | |
| 1987 | 29 | 34 | 37=100 | +5% |
| 1989 | 33 | 33 | 34=100 | +0% |
| 1990 | 31 | 33 | 36=100 | +2% |
| 1991 | 31 | 32 | 37=100 | +1% |
| 1992 | 28 | 33 | 39=100 | +5% |
| 1993 | 27 | 34 | 39=100 | +7% |
| 1994 | 30 | 32 | 38=100 | +2% |
| 1995 | 32 | 30 | 38=100 | -2% |
| 1996 | 29 | 33 | 38=100 | +4% |
| 1997 | 28 | 33 | 39=100 | +5% |
| 1998 | 28 | 33 | 39=100 | +5% |
| 1999 | 27 | 33 | 40=100 | +6% |
| 2000 | 28 | 33 | 39=100 | +5% |
| 2001 | 29 | 34 | 37=100 | +6% |
| Pre 9/11 | 28 | 35 | 37=100 | +5% |
| Post 9/11 | 31 | 32 | 37=100 | +1% |
| 2002 | 30 | 31 | 39=100 | +1% |
| 2003 | 30 | 31 | 39=100 | +1% |
| 2004* | 29 | 33 | 38=100 | +4% |
| AVG | 27.73 | 29.26 | 37.89 | +3.53% |
So you see, the average Democratic Voter Identification, the amount of voters who at an exit poll identify as a Democrat, has a 17 year average of 29.26% of the population. Or just above a 3.5% margin over Republicans. This is not disputed. Show me a pollster who disputes this, and I will show you real bias.
So why did Gallup commission a poll that didn't even come close to matching what actually exists in real life? That I don't know. I don't know if it is actual malicious bias, or if their internal algorithm is just wrong. Either way, we can never listen to Gallup polls as being authentic again.
Additionally, I will say it again, you cannot use Likely Voters (LV) until the week of the election. This is because Likely Voters fluctuate daily due to momentum of each campaign. Likely voters measures momentum, which is important to know on the eve of the election. A more reliable poll sample would be Registered Voters (RV), which fluctuate less with the winds of change.
Furthermore, what is interesting about this poll, is that Bush's lead over Kerry has narrowed between the R7% biased poll of two weeks ago and yesterday's R12% biased poll. That means that Kerry has effectively taken the lead is the polls were normalized for actual party identification.
Trackback Pings
TrackBack URL for this entry, Gallup Poll now officially laughingstock:
- September 29, 2004 · Grubbykid.com :: Words - More on the Gallup Poll
Comments
Post a comment
This is the permanent home of Gallup Poll now officially laughingstock. I wrote this post at 13:24 on September 28, 2004. This post is part of grubbykid.com, a weblog. If you liked this entry, why don't you read some other posts such as Daily Show viewers smarter than O'Reilly Factor viewers or Bush Would Use Power of Persuasion to Raise Oil Supply? Or you could go to the site archives or return home. All are good choices.
