Tuesday :: Sep 21, 2004

Gallup's Iowa Poll From Yesterday Is Suspect Also

by Steve

Stop me if you’ve heard this one before. Gallup and USA Today have announced the results of Gallup’s latest poll from Iowa, wherein Bush allegedly leads Kerry by six points (50%-44%) amongst likely voters in the state. The poll included 631 total respondents, but the weighted base was reduced down to 554.

Here’s the party ID composition of this sample from Gallup yesterday:

Weighted Base: 554

GOP: 218 (39.35%)
Dem: 182 (32.8%)
Ind: 153 (27.6%)

Look familiar? It’s roughly the same proportion of Republicans to Democrats that Gallup uses when they screen for likely voters in their national samples, as I discovered last week. Gallup is producing polls at the national level that overstate GOP representation from the last two presidential elections by approximately five percent and understate Democratic representation by nearly six percent. And now it appears they are doing so with their state polls as well. How do I know?

Because according to the Iowa exit polls from the 2000 race, these were the results by party ID:

GOP: 33%
Dem: 35%
Ind: 31%

So, we have a case where Gallup is using a sample that produces a GOP advantage of at least 6% over the Democrats when in fact the Democrats had a 2% advantage in that state in 2000. And in at least two states (Wisconsin and Iowa) these polls are being used by the media to claim that Bush is moving ahead of Kerry in key battleground states.

Again, how likely is it that the GOP will have a 6% edge over the Democrats in a contested battleground state that went for the Democrats last time by 2%?

My simple message is this: don’t assume that Bush is moving ahead in these states based simply on Gallup polls.

Steve :: 3:02 PM :: Comments (15) :: Digg It!