MyDD has an excellent analysis of the Newsweek poll--sample bias

[kstewart33]

From MyDD:

Democratic self-identifiers come in at around 1/3 of the electorate, with Republican self-identifiers just slightly below that. If we can assume that those who identify as none of the three choices split proportionally to the rest of the population (not a stretch by any means), Harris would come out to 39% DNC, 33% RNC and 28% indy. For NAES, the numbers would be 38% DNC, 36% RNC, and 26% indy. Considering this, much of the bounce from the Newsweek poll is generated by poorly weighted Party ID numbers.

Party ID Bush Kerry Und/Oth
Republican 38% 94 4 2
Democrat 31% 14 82 4
Independent 31% 45 40 15
Total 100% 52 41 7

However, if the sample was properly weighted, here are what the results would look like:
Party ID Bush Kerry Und/Oth
Republican 35% 94 4 2
Democrat 38% 14 82 4
Independent 27% 45 40 15
Total 100% 50 43 6

(snip)
In reality, in the Newsweek poll Kerry is only down by seven in a three way trial heat, and thus is probably only down by 5 or 6 in a two way trial heat.

Link: http://www.mydd.com/

[Nancy Waterman]

1. What about theTime poll

Was there the same over Repub bias there?

[GaryL]

2. Yep.

However the excuse they use is pugs a "more motivated" and thus "deserve" a greater "likely voter" count. Assholes.