NEW COMPOSITE POLL
In the new poll this week, there is little change from last week. I didn't use the Marist poll because (a) it came out later in the day, and (b) it wouldn't have really changed the result. I am now using Rasmussen's poll numbers and doing a weighted average of the state-by-state polling, giving greater weight to more recent polls. I'm also including a lot more data on the composite page. Click here for the full analysis.

The numbers for both Kerry and Bush are up this week, though the margin is about the same. There still aren't a lot of polls coming out weekly, so things should settle down when more polls are released weekly as we get closer to the election. Although the EV prediction is similar to last week, with Kerry picking up Iowa, many states are tighter than last week and there are more battlegrounds.

in future weeks I'm going to rejigger the way I allocate the undecided vote, based on how they broke in past elections with incumbents running. Dick Morris suggests that they break 85/15 for the challenger, but my analysis thus far does not verify his conclusions.