September 27, 2007
POLL: Gallup National Democratic Primary and Clinton
New analysis aggregated from the four most recent Gallup national surveys of 1,989 Democrats and those who lean Democratic (conducted 8/3 through 9/16) finds:
- Sen. Hillary Clinton leads Sen. Barack Obama (47% to 25%) in a national primary while former Sen. John Edwards trails at 13%. All other candidates average at less than five percent each.
- "Clinton holds a commanding lead among nearly every major subgroup of potential Democratic primary voters."
- A few "weak links" for Clinton: Among those in households with an annual income of $75k or more, she leads Obama 37% to 30%. Among Men aged 18-49 she narrowly leads Obama 39% to 34%.
By Eric Dienstfrey on September 27, 2007 6:34 PM | Permalink
Comments
Correction. I said the poll was conducted in 16 days. It's 13.
I think a poll conducted over 13 days shows consistency in the polled group. Events happened during that 13 day span and Clinton maintained her lead essentially unchanged.
Whatever Obama and Edwards have done has not changed the landscape of this election.
Two thoughts: First, we are talking about four surveys conducted over a 44 day span. -- that's August 3 through September 16.
Second, if you click through to the Gallup analysis, you'll see a remarkable consistency in the overall vote preference on those four surveys. So while Andrew is right that voters' minds could change over that span of time, the evidence from these surveys shows that they have not on the measure in question.
So it's 44 days, not 3 as I believed. Wow. I definitely not trust this methodology.
What is the disadvantage of simply using the Sept. 14-16 period, compared to averaging all those semi-ancient surveys together?
Since August 3rd we have had the David Petraeus report, the Larry Craig incident, the Hsu scandal, the Moveon ad news, Giuliani staging a phoney phone call from his wife, Fred Thompson announcing his candidacy, etc.
Even if we assume Gallup got lucky and voters didn't care much about this events, professional pollsters should not rely on luck.
I assume this is why most other pollsters conduct polls in 1, 2, 3, or 4 day periods.
Andrew:
The point of this release was to analyze the results of just one question -- the Democratic primary trial heat -- among subgroups of Democrats. If they had just used the 9/14-16 data, the subgroups for gender, race, age, etc would have been quite small and less reliable.
They looked at the overall numbers across the four polls and saw no change. So they concluded that it would be reasonable to roll together the data in this instance for purposes of subgroup analysis.
Ok, that makes more sense, although the effectiveness of this approach is potentially limited, even for a primary, because it can only be used when no major political events have changed the landscape.
A poll such as this one, conducted during a span of 16 days cannot be a current reflection of voter preference. Too many events that change voters' minds happen on a daily and weekly basis.
Posted on September 27, 2007 6:52 PM