« Persuadable Voters Still Not Persuaded | Main | Kerry Seizes Lead in New National Poll »

White House Race Narrows in New National Poll

A head-to-head poll of nation-wide LV's by Greenberg Quinlan Rosner Research conducted 9/12-14 for Democracy Corps has Bush ahead of Kerry 49-48 percent with 1 percent going for other.


It looks to me like they have their own totals incorrect. When asked who you would vote for (or something similar),
46% said Kerry
1% said Lean Kerry
48% said Bush
1% said Lean Bush.

Wouldn't that mean the poll shows Bush ahead 49 to 47, as opposed to 49 to 48?

Presumably roundoff error is playing a role here.

timshel, They likely rounded off the numbers. If the Kerry number is 46.4 and the leaning Kerry is 1.4, that adds up to 47.8, which rounds off to 48.

Here's more good polling news: Harris has Kerry leading by among LVs, and Bush's reelect number is just 45 percent. I read this on the Wall Street Journal site.

My post should say Kerry is leading by 1 among LVs.

What I don't understand is how many news magazines are continuing to use poll data taken during the Republican convention. Today's TIME had the same Sept. 7 poll that most newspapers stopped citing over a week ago.

Seems to me like in the information age the print media is always a bit behind the curve.

Page 19 of this new poll shows that of those who were polled, 43 percent voted for Gore in 2000, and 51 percent voted for Bush. Since we all remember that more people voted for Gore in 2000, there seems to be a bias in the polling sample towards Bush. There is no indication that the polling percentages were weighted to account for this, but it's hard to believe that wouldn't be taken into account.

The biggest difference between now and last month is that all polls switched to using LVs, which exaggerated bush bounce. Conversely it will also underestimate the drop in the bounce.

Paul, what new poll are you referring to? The Democracy Corps survey? I know that the Time magazine poll had an similar unrepresentative majority of Bush 2000 voters.

Never mind. I just found it's the Democracy Corps poll.

Remember that when people are asked who they voted for in 2000, it's quite likely that some will be in error. Perhaps they did not actually vote, but want to identify with a winner, so they say they voted for Bush. Or perhaps they want to convince themselves they voted for a winner and so say they voted for Bush. Evidence of differential reported voting patterns in 2000 does not necessarily mean you're looking at a nonrepresentative sample.

Yeah, tony, that occurred to me, too (the most famous example is the percentage of people who, post-assassination, said that they had voted for Kennedy--it was a huge percentage). In this case, the other numbers are not representative, either. 80 percent white, 10 percent black, 7 percent hispanic, whereas those last two numbers were in the 12-13 percent range in the 2000 census. What are the demographics of people who no longer have land lines, but only cell phones? Who will they vote for? Does any pollster know?

Any body have suggestions though on why Rasmussen's rolling daily poll suddenly showed what had been a Bush lead of one percent or less for several days jumping to about five points today? This is a three day rolling poll, so if the numbers are remotely correct, something weird happened. I'd love for somebody to 'explain' this since Rasmussen has not seemed off base before. T.J.

Rasmussen just switch from a 3 day rolling average to a 7 day or 14 day rolling average. This could explain the sudden jump in the number since it would include some of the early post-convention data If true it will smooth out in about 4-7 days.

Probably a sampling error. When you take a poll every day, you wind up with a bad sample now and then. When looking at a tracking poll, the important thing is the trend, not the daily number. The trend for Bush has been downward since the end of the convention and Kerry's trend has been upward since then. When the bad sample rolls off, the numbers will narrow again.

The difference is there in both 3-day and 7-day tracking data. Never mind! Strange though how the spread is similar in both 3 day (4.6) and 7 day (3.7). The one day jump must have been enormous to add 2% difference to the previous 7 day average.