RE: 'tied with Nader'
Yes...i skimmed it and saw this and didnt note the 'battleground' part.
In the battleground states, it was Bush, 45 percent; Kerry, 45 percent; Nader, 5 percent; and undecided, 5 percent.
also of note is:
In 17 states considered key battlegrounds, the Massachusetts Democrat was favored by 47 percent of voters to 44 percent for Bush, again with 9 percent of voters undecided.
These results are similiar to the recent ABC/WP poll that showed Bush winning Nationally, but Kerry winning in BG states.
So another headline for this poll is: Kerry +3 in Battleground States
Re: Odd Battleground Numbers
ABC News "kinda" does their polls like Zogby. (and kinda not, Zogby is his own unique animal)
Both Zogby and ABC break the population down into demographic groups (I think ABC uses 78 categories, but that is from memory, so it may be wrong) and then have a "quota" system when they collect data - ie they need this many rural white people, this many city blacks, so many with kids, etc...
While this, at least in theory, produces a good national sample, the regional samples can be really, really weird.
You tend to get your "easy" quotas (Elderly, women, unemployed) in the early time zones, and then the "hard" quotas (Empty nesters, etc) in the later time zones, which really skews the subsamples.
For example in the last Zogby, Bush was up only 4% in the south, while being up 6% in the Central/Great Lakes area. - Clearly a bit off I am sure you will agree...
Finally, in the ABC poll, they talked to +/- 1000 people which equates to about 400 in the "Battlegrounds" ie +/- 5% or so.
Regarding the Marist Poll, I do not know enough about their inner workings to even suggest why... If I see any breakouts, I will let you know
The polls (aproximately) break into two groups..
Among those who do not weight/construct samples, the regional breakouts should, bearing in mind the big MOEs of subsamples, be fairly good.
In this category would be:
Fox (kinda sorta, they semi-constrain their samples)
CBS (although CBS is just so fundementally bad, I'd just toss the whole poll anyway)
If any of these 4 has "crazy" subsamples, toss the whole poll, as they clearly just had a bad sample (If Gallup gets weird internals they just throw the poll away and do another one BTW)
Among those who do weight, regional breakouts are very. very iffy..
In this category would be:
Neither the "weighters" or the "non-weighters" is "better" there are excellent firms in either camp, they just have different characteristics. And one of the characteristics is that the non-weighters can (subject to huge MOEs) produce semi-useful subgroups, while the weighters/constructive samplers cannot as a rule.
Hope that is a helpful explanation!