Romney in Iowa: Don't Believe the Hype
by Ryan Sager
Tue, 3 Apr 2007 at 4:23 PM
updated Tue, 3 Apr 2007 at 4:37 PM
Romneyites are pushing a poll today showing their boy breaking double digits (how exciting for them!) in Iowa. It's from the University of Iowa and shows McCain at 21%, Giuliani at 20%, and Romney at 17% among likely Republican caucus goers.
There are a number of reasons not to take this poll too seriously.
First off, it doesn't show Mitt moving, as there's no previous UI poll to which to compare it. Secondly, the sample should make anyone queasy: it skews far too female (62% -- remember the 2004 exit polls?), and it's far too small in the subgroups (we're talking about 178 GOP caucus goers).
This ARG poll, meanwhile — conducted in a timeframe that overlaps with the UI poll — had 600 likely Republican caucus goers. It pegged Romney at 10% (Giuliani and McCain were tied at 29%). Unless Mr. Romney's had a huge uptick just in the last week, all the UI poll shows us is differences in survey methods.
David Redlawsk, who conducted the UI survey, told me over the phone a few minutes ago that they didn't release a weighted result (to make up for the gender imbalance) because nothing important changed when they ran the numbers. Still, a sample that small and skewed is suspect by default.
We'll watch to see if the Romney Iowa boomlet shows up anywhere else.
UPDATE: Mr. Redlawsk writes in with this clarification in the comments. I'm moving it up here:
Ryan is right to suggest that one poll doesn't tell the whole story. And no doubt our poll at the University of Iowa has a relatively small sample of likely caucus goers. A reason for that is that the subsample was part of a larger poll of registered voters - but at best about 20% of registered voters are likely to caucus. This leads to small subsamples, unless special efforts are made that we did not have the resources to do this time.
The one clarification I'd like to make is my fault, not Ryan's. We released the overall gender breakdown of the full sample (n=1290) which is heavily skewed female. However, I failed to note that the Republican likely caucus goer subsample is actually heavily MALE - 64.2%, FEMALE 35.4% (The Democratic Caucus sample is 51% Female, 49% male.)
It turns out that our Republican sample gender breakdown actually tracks with what little data there is about actual caucus attendees. Researchers at the University of Northern Iowa have administered surveys at actual caucuses over many years. In their results, just over 59% of Republican caucus goers are in fact male!
So while it is true the sample is small and is only one point in time, it turns out that we are not all that far off on gender, so I have a reasonable level of comfort in our results, at least on that score.
Dave Redlawsk
University of Iowa
Latest Politics Homepage
|
Print
Send
RSS
