The Latest on the Polls

Today’s Rasmussen tracking poll is the first in which all or substantially all of the interviewing followed the close of the Republican convention. Unfortunately, it shows President Bush ahead by only one point.
Scott Rasmussen says that later today, he will post an analysis of why his poll differs sharply from the Time and Newsweek polls that give Bush an 11-point lead. I’d encourage you to read it if you are interested in learning more about polling. I’m glad that the news is full of talk about Bush’s bounce and the Kerry campaign’s shakeup, but I think Rasmussen’s polling is more reliable.
On a positive note, Rasmussen notes that Bush and Kerry are currently tied in Minnesota, which is consistent with what I’ve been told about non-public poll results, and that Kerry is ahead in California by only 8 points.
BIG TRUNK adds: Reader Rick Reese provides cautionary notes regarding Rasmussen. Reese writes:

Actually, Rasmussen does not have a very good track record.

Reese quotes a post from Free Republic as follows:

Here are Rasmussen polls from 2000:
Oregon – They had Bush by 5, Gore won by 1
Arkansas – They had Gore by 2, Bush won by 6
Maryland – They had Gore by 1, Gore won by 17
Out of 6 polling services who did numerous state polls in 2000 (including Zogby, Research 2000, Mason/Dixon, FoxNews, ARG), Rasmussen ranked 6th, Dead Last in accuracy.

See also D.J. Drummond’s analysis of recent polling data: “First look at a big bounce?”
UPDATE: Although I don’t see it on his site, apparently Scott Rasmussen did release his analysis of why his current results are so different from Time’s and Newsweek’s. Rasmussen says that he got what he considers “rogue” data on Saturday. Saturday’s results showed Kerry with a big lead, while all of the surrounding days show a healthy lead for Bush. (This is consistent with the conventional wisdom that weekend polling favors Democrats, because Republicans go out more.) Rasmussen says that he thinks President Bush actually has a four or five point lead.
Rasmussen thinks that the Time and Newsweek polls oversampled Republicans, and provides a plausible statistical analysis of why that would be true.
Meanwhile, the new Gallup poll has Bush with a seven-point lead.
All of which I take to be good news.


Books to read from Power Line