WASHINGTON

                              KERRY BUSH NADER MOE

Strategic Vision  9/20 - 9/22  46    45    2    3   (Kerry by 1)

Survey USA        9/19 - 9/21  51    46    -    4   (Kerry by 5)

Ipsos Pub. Aff.   9/17 - 9/20  51    42    2    4.9 (Kerry by 9)

Elway Poll        9/17 - 9/19  52    38    -    5   (Kerry by 14)

CALIFORNIA

                              KERRY BUSH NADER MOE

Strategic Vision  9/20 - 9/22  48    43    -    3   (Kerry by 5)

Los Angeles Times 9/17 - 9/21  53    40    2    3   (Kerry by 13)

Pub. Policy Inst  9/12 - 9/19  51    39    -    3   (Kerry by 12

I think, as I've expressed before, that what you see is a deliberate attempt to undermine Kerry's public position by shaving off points from his poll results wherever and whenever possible -- always being careful never to go too far and present results that no one will believe. The margin of error gives a fair amount of wiggle room for these guys to play within, and I think they're using it for all it's worth.

It would really be best if people dropped SV's results from their poll-driven Electoral College estimations, because they're not trustworthy, in my humble opinion.

For more, see this anlysis and the further links in that post.


Update (9/28): In an e-mail to me, TruthIsAll had some interesting things to say:

I have been posting on just this topic at DU [Democratic Underground].

On the state level, a comparison of results from the model.

On the national level, to distinguish between National pollsters, I have created a separate group of 8 Independent pollsters, who are part of the existing full group of 18 corporate and independents.

Is it just a coincidence that the Independent 8 National Group has the race tied at 46%, whereas the full 18 Group has Bush leading by 2%? This means the corporates have Bush leading by an average of 4%.

The effect on the probability of a Kerry win (defined as the probability of getting a majority of the popular vote) based on these two National polling groups is very interesting.

In an unbelievable coincidence, they are exactly reversed as of today.

For the 8 independents: Kerry has a 92.38% win probability.
For the full group of 18, Bush has a 92.38% win probability.

The State model, based on Monte Carlo Simulation, gives Kerry a 91% probability. So the State model confirms the Independents National model.

The numbers are quite interesting:

State Model:
Kerry current poll weighted average is 46.49%, projected to 50.85% with a win probability of 91.0%.

National 8 model:
Kerry current poll average is 46%, projected to 50.80%, with a win probability of 92.38%.

This is a key point: Both models use totally different methods to compute probabilities and derive Kerry's national popular forecast.

The National model applies the tight MoE for the 8 polls (8,000 respondents) to the projected Kerry % to derive his probability.

The State model uses the results of 5000 simulations based on the latest weighted state polls (mostly Zogby and ARG) to calculate the probability (the number of Kerry wins divided by 5000).

The race is tied if one looks at the current national 8 average poll and the weighted state polls.

But it's NOT really tied, since Kerry will get the bulk of the remaining undecided and other 3rd party voters, not to mention the effects of the unknowable and potentially massive democratic turnout.

Kerry is on the verge of closing the deal.

TruthIsAll