I see several problems with the Newsweek poll.
#1. It was conducted almost entirely on Friday and Saturday morning. It didn't poll the Eastern and Central time zones on Thursday night. Weekend polls like this usually undersample Republicans.
#2. It has more Democrats than Republicans. I am not certain about the over-under sampling of partisans. I tend to think if the poll is properly conducted, you need to look at party ID as a dependent variable, one that changes with ups and downs. Now -- are we to believe that in the last month there has been more than a 8% net switch in partisan identification? That is what the Newsweek poll from 10/2 stacks up with the Newsweek poll of 9/10.
#3. It is not problematic to me that it is 52% women and 48% men. What I find problematic is that the *men* who were surveyed support Kerry 47% to 45%. Not even Bob Dole did this poorly in 1996. This indicates that *they undersampled Republican men.* This would, incidentally, make sense in light of the fact that almost all of the sample was done on Friday. What are large swaths of Republican men doing on Friday nights in those red states in the fall? A. During the day they are working; B. During the evening they are at HIGH SCHOOL FOOTBALL GAMES!
(This, my friends, is the reason Kerry is in the lead. He has an insignificant lead among women and Independents...but he also has a similar lead among men. This is unheard of.)
Newsweek usually rushes polls out. I think they do it not so much to get a sense of the electorate, but to make news. Rember that Newsweek is the only poll with the honor of:
A. Showing Bob Dole down by 2% in 1996.
B. Showing Walter Mondale beating Reagan in 1984.
It is not because they are biased, but because they rush their polls out. The people who are sitting at home on Friday day and Friday evening are not a representative sample of the general American population.
All polls need to be filtered by common sense. This Newsweek poll is like that Pew polls from September. The first showed Bush up by >10, the second showed Bush up by 1, the third showed Bush up by >5. Now, do we really, honestly, truly think that the electorate has bounced around so much?
Similarly, with this Newsweek poll. Do we really, honestly, truly think that Bush is losing men and almost tying women? Do we really, honestly, truly think that the dynamic of the race has shifted by *eight points* since September 10?
If so, I have a bridge I'd like to sell ya! :-)
Great analysis, thanks.
Newsweek Partisan Shift = 11.41%!!!
I can't say it is more or less of a shift in favor of Kerry because the new poll doesn't have the breakdown of what support each candidate is getting from the other party. But it is clear that the partisan make-up of the last poll compared to this poll is dramatically different and almost certainly the source of the shift in support for Kerry.
Last Newsweek Poll (http://biz.yahoo.com/prnews/040911/nysa006_1.html):
Republicans: 391 (40.73% of total sample)
Democrats: 300 (31.25% of total sample)
Independents: 270 (28.13% of total sample)
TOTAL: 960
New Newsweek Poll (http://www.prnewswire.com/cgi-bin/stories.pl?ACCT=104&STORY=/www/story/10-02-2004/0002263797&EDATE=):
Republicans: 345 (34.95% of total sample - 5.78% fewer Republicans)
Democrats: 364 (36.88% of total sample - 5.63% more Democrats)
Independents: 278 (28.17% of total sample - .04% more Indpendents)
TOTAL: 987