Tom,

I realize "there are too many Dems" is a standard way for the right-blogosphere to dismiss poll results. As I posted and you read before, that's usually not right. I'll summarize once again for your convenience:

1. Party ID is a *dependent *variable in opinion surveys, not an unchangable demographic characteristic. Reported party ID varies as attitudes towards the two parties varies, yes, even if people don't go out and change their actual registration.

2. If you read the actual questions asked, you'll see that AP/Ipsos asks which party "you feel closer too", not "what is your official registration". This asks people to report the party they currently happen to like, rather than reporting their actual registration. This makes it even less likely that the numbers will match registration data.

3. AP/Ipsos pushes Independents to choose a party. Again, this guarrantees that the numbers will not match registration. The fact that Independents now break much more towards Dems than they did two years ago is a sign of an interesting political shift, completely consistent with other data, rather than a sign of bias. As for that other data, on surveys that ask for actual ID and don't push Indepenents into a party, Bush approval among Independents is much closer to Bush's rankings among Dems than among Republicans, and likewise on many specific policy areas, Independnts are closer to Dems. So I'm not surprised that when you push Inds to choose a party, more choose Dems than the GOP right now.

You are, of course, free to believe it's due to outright fraud. Just as you are free to be surprised by election results when you dismiss these polls.