Quinnipiac to Steve Schale: Please get your facts straight
We noted yesterday Democratic strategist Steve Schale's criticism of the latest Quinnipiac poll. Today Doug Schwartz, director of the Q poll, responds in a blog post, explaing why Quinnipiac prefers to rely on people's self-described party identification instead of weighting to self-described party registration:
Yesterday’s Quinnipiac University Poll on the presidential race in Florida was criticized by Democratic consultant Steve Schale for the partisan and racial composition of its sample. The charges were repeated in a blog by the Tampa Bay Times’ Adam Smith. The biggest problem with this analysis is that Schale got the Quinnipiac numbers wrong!
“The Q poll, which gave Mitt Romney a 6 point lead, weighed out at 37% Republican, 29% Democratic and 29% Independent. It also landed at over 80% white, 8% Hispanic and 7% African America and Caribbean American.
Those were NOT our numbers. Our weighted sample was 34% Republican, 31% Democrat and 29% Independent. It also was 69% white, 14% Hispanic, and 12% black.
This was not the only time Schale misstated the Quinnipiac numbers.
In fact, if you go back to the last Q poll, which had the race 44-43 Romney earlier this month, that poll also had a bizarre electorate make-up of about 32R-30D-28NPA, again a scenario that is simply not going to happen on Election Day….”
Schale was wrong again.
Our last Florida poll released on May 3rd was 28% Republican, 31% Democrat, and 37% Independent.
The other problem with Schale’s analysis is that he uses party registration figures to suggest that our party identification numbers are not accurate. This is comparing apples to oranges. Party registration and party identification are not the same thing. Most major pollsters, such as CBS, ABC, Gallup, and Pew rely on party identification, not party registration, when doing general election polling. The reason is that party identification is considered a better measure of voter attitudes toward the parties than party registration. Some people register with one party and may change their attitude about that party but not bother to change their registration. Other people may simply have forgotten their party registration. When trying to predict voting behavior, party identification is a much better predictor than party registration.
For a media explanation of the differences please see the following:
For additional information on party identification in election polling please see the following:
The wording for our party identification question, as well as the results, can be found in our demographics document that is attached to every Quinnipiac University Poll release. Here is the link to our most recent Florida poll. A link to the demographic data can be found at the bottom of the page beneath the tables.
Based on the Republican and Democratic pollsters Marc Caputo and I have spoken to, most weight based on voter registration, including the Times/Herald non-partisan pollster Brad Coker of Mason Dixon. Democrat Dave Beattie noted that in other states he often relies on party identification rather than registration, but Florida is blessed with excellent voter registration data that he said makes registration a more reliable measure for weighting.
"I think Quinnipiac does a pretty good job," Beattie said, "but I wish they'd spend as much time thinking how they could do it better as they do being defensive about it."