Polls at their best are useful only as trend-indicators, not as absolute data.
However, America has once again exhibited its "irrational exuberance" for something and seems to be accepting them as gospel, so they need to be looked at a bit more critically.
I'll start with the most recent ABC poll, which asked: "Who, in your opinion, won the [vice-presidential] debate?" The answers were:
[ABC News Poll. Oct. 5, 2004. N=509 registered voters nationwide. Fieldwork by TNS.]
| Cheney | Edwards | Tie | |
| Responses | 43% | 35% | 19% |
The difficulty with that is the party affiliations of the respondents:
| Democrats | Republicans | Independents | |
| Party ID of respondents | 31% | 38% | 27% |
That's a problem because the sample does not represent the make-up of the national electorate, which is:
| Democrats | Republicans | |
| Party ID of registered voters | 33% | 29% |
Clearly, the results are skewed by using more Republicans than Democrats when the voter population breaks the other way.
Taking that into account, a more accurate number for Cheney might be around 35% (20% less) and the real result is about a dead heat.
It's interesting that TNS Intersearch, the organization doing the field work on this poll, on their site loudly proclaims:
one + one = three
They're right.
And they're not the only ones doing it.
Gallup polls have been around forever and you'd expect them to be honest, but they do the same thing:
The Gallup Poll, despite its reputation, assumes that this November 40% of those turning out to vote will be Republicans, and only 33% will be Democrat. Gallup just confirmed for me that this is the same sampling methodology they have been using this whole election season, for all their national and state polls.
Continued on the next page Page 1 — Page 2
Article comments
1 - Voxxy
Given the spanking that Cheney delivered last night, the accurate figure should be 90-10 for Cheney. In two debates, Kerry and Edwards have said nothing.
2 - Cap'n Ken
I don't doubt that pollsters - like the media - bring biases into their work.
But on the whole party affiliation thing, I don't think you can make a valid comparison with those numbers. The way a pollster asks someone about their party affiliation won't necessarily match up with a figure about the percentage of registered democrats or republicans.
I assume your registration data comes from party affiliations in voter registration records. Some states require you to register as a member of a party, some don't. Some let you say "independent", some don't. The fact that your Party ID table only adds up to 62% certainly implies it doesn't give a complete picture. Perhaps the states that require party ID in registration lean Democrat, or maybe they lean Republican.
I think having a pollster ask respondents what party they are affiliated with is a more valid representation of that sample's leanings.
That's not to say Gallup isn't out to find more Republican voters.
3 - Hal Pawluk
More Democrats than Republicans voted in the last two presidential elections.
To make any sort of reasonable projection about the next election from polls, you need more Democrats than Republicans in the sampled universe.
4 - joe
i'm doing a project on slanted polls any suggestions?