October 03, 2004
The Newsweek Poll
Newsweek Magazine released a survey last night, "the first national telephone poll using a fresh sample” of all registered voters nationally since Thursday night’s debate. The results, if taken at face value, are highly encouraging for the Democrats, showing Kerry two points ahead of Bush (49% to 47%) in a three-way match-up that includes Ralph Nader (at 2%), though the story appropriately characterizes the result as a "statistical tie.” Without Nader included in the question, the survey puts Kerry ahead by three points (49% to 46%).
We have good reason to be cautious, however, as Newsweek did most of its interviews on Friday night and Saturday afternoon. According to the Newsweek release, it conducted the survey "after the debate ended” with interviewing done Thursday through Saturday. Since the debate ended at 10:30 p.m. Eastern time, I assume that Thursday night interviewing was limited to the West Coast and very light elsewhere.
Because so many adults are away from home on Friday nights and Saturdays, very survey organizations conduct polls on only Fridays and Saturdays. In my experience, weekend-only interviewing yields respondents that are more aware of current events and political figures even after demographic weighting.
As some may recall, Newsweek was similarly out front with a poll done immediately after the Republican convention (Thursday and Friday, September 2-3). They had Bush ahead among registered voters by a slightly wider margin (52% to 41%) than other surveys done during the following week.
Also, note that Kerry's six point gain (from 43% to 49%) since the September 9-10 looks to be right on the edge of statistical signifance given the reported +/-4% sampling error. Perhaps that's why the Newsweek poll story omits any reference to the September 9-10 survey and instead hypes the more dramatic and significant (though misleading) comparison to the eleven point Bush lead (52% to 41%) on Sept 2-3. The Newsweek release and Reuters story included the data from 9/9-10.
To be clear, I am urging caution about these results, not disbelief. We can have much more confidence when we see results less reliant on Friday-Saturday interviewing. I’m assuming many, many new polls will follow in the next 72 hours.
Sep 2-3: 374 Republicans, 303 Democrats, 300 independents
Sep 9-10: 391 Republicans, 300 Democrats, 270 independents
Sep 30-Oct 2: 345 Republicans, 364 Democrats, 278 independents
I could understand large swings in party ID if these were likely voter polls. Republicans were probably more apt to respond after their convention, and Democrats after their candidate won the debate.
But the Newsweek polls survey registered voters, so this looks very strange.
Posted by: Tom L | Oct 3, 2004 1:51:07 PM
Sorry, I didn't express that very well. I meant that with likely voter polls, party ID might conceivably skew because one party or other is actually more likely to vote. Non-response error could obviously affect both types of polls, as could a real shift in partisan alignment. But these changes seem large no matter what the explanation.
Posted by: Tom L | Oct 3, 2004 2:01:21 PM
I had always thought, in my nonprofessional ignorance, that polls of registered voters were less accurate than those of likely voters. I have read during this election cycle that this A) depends on how you define a likely voter, and B) even then is not always true. I would be very interested on the Mystery Pollster's observations on this.
If, as I used to believe, polls of registered voters are more suspect, I would assume that someone at Newsweek, et. al, would know this. Why then would they use the registered voter standard?
Posted by: Assistant Village Idiot | Oct 3, 2004 2:52:46 PM
Mark, this may be a silly question, but given that there was a 12-point shift in party ID toward Democrats relative to the preceding Newsweek poll, would it have been reasonable to expect a 12-point shift in support for Kerry (as opposed to the actual 8-point shift)? Thanks.
Posted by: Larry Jones | Oct 3, 2004 9:33:24 PM
Slightly off topic, but:
Do you have any thoughts on the Dispatch poll, which is done by mail?
Posted by: Dr. X | Oct 4, 2004 3:18:53 AM
I think non-weekend polling might be becoming a rarity for the next two weeks.
There are too many swing events going on during the week to jam poll around. We have three debates left, tomorrow, Friday and next Wednesday. Until that last debate, I think most poll companies will try to hit the calm weekend points and not try to slip a weekday poll in.
When would you do a weekday one? This week might not be too bad, because I don't think the VP debate will mean much over all. I suppose you could do one Sunday night - Tuesday, but it would be a bit of a wasted effort, coming just before the last debate.
I don't think we'll see the first solid non-weekend polls until the 21st or 22nd, and that's just a hop and skip before the election.
Posted by: Ann | Oct 5, 2004 12:05:36 AM
The comments to this entry are closed.