Rasmussen Reports 5-6-04 Kerry 45 Bush 45

[The Vorlon]

Also to Vorlon, I remeber you saying you did expect Bush to pull within one or two in today's sample. With Bush pulling even, did Bush poll ahead of Kerry last night? Or was it simply the dropping of Sunday night's results? I'm sure everyone agrees with me, but I appreciate your analysis and enjoy hearing it.

I posted a real "quickie" take on the Gallup FYI.

Regarding last night's Rassmussen sample:

Actually, Bush is still "down" 45.3 to 44.7% although certainly .6% is statistically meaningless.

You notice I stopped posting a daily breakout of Rasmussen a few weeks back.

I had a faily long set of email exchanges with Scott Rasmussen on this matter because I was having trouble getting his daily results to reconcile with his 3 day rolls.

Rasmussen does not "exactly" produce a one day sample.  He does drop off the data from 3 days ago and replaces it with the latest one day set of data, but then he reweights the entire 3 day data set  - there is no "one day" total per se.

I have reworked my daily breakout of Rasmussen by recursively reweighting the individual days in a way Scott intimated would be appropriate, and based on that, yes indeed Bush "won" last nights sample by about 2.7% (In a recursive weight I can calculate the "gap" between the candidates but not the exact percentages)

Given the big bounces that happen in a smallish one day sample, I urge real caution on reading anything into this one day result.

Recursively pulling the "gap" is even more likely to be subject to big statistical changes. - I can say Bush "won" by 2.7% last night with greater accuracy than using dice and tarrot cards - but only marginally so..

[The Vorlon]

Is it possible that Rasmussen is using a high number of DEMS and GOP and few IND? If you assume that Rasmussen is using a ratio of say 41 DEM/38 GOP/21 IND, and if each candidate is maintaining the same level of support from their base, the primary thing that is swinging the numbers from day to day is the IND vote. I would also assume that a large portion of the undecideds are coming out of the IND portion, further reducing the amount that could swing from one candidate to the other.

So taking the 21% IND, reducing it by 4% for undecideds leaves 17% available to swing.  This means that the IND would have to swing more than 60-40 to one candidate or the other to get the topline results to be outside of the 43-48 range.  

So, if it is true that Rassmussen is using a low number of IND, how can the results he gives really change unless one of the candidates starts to lose their base?

Scott Rasmussen has been rather coy on his exact weighting formula.  Pollsters guard things like this closer than Col. Sanders protects his 11 different Herbs & Spices

Based on my emails with Scott it "think" that I have "read between the lines" that he is using 38/35/27 Dem/GOP/Ind. which explains and supports your theory.

Tracking his data versus his own "Generic Congressional" poll I am "fairly" sure he is weighting at +3 for the Dems, but I would not guarantee it.

Boss Tweed has emailed Scott point blank asking him what his weighting formula is.  I hope he gets the striaght answer I did not...

[The Vorlon]

Coming Today...

3:00 p.m. Eastern Florida Tracking Update for Premium Members

You think these numbers will leak out? Or should we all chip in $1 dollar to Vorlon and raise the one time fee of $49?

Rough and dirty....

Michigan 17 Evs - 2 bonus EVs = 15
Ohio 20 Evs - 2 bonus EVs = 18
Pennsylvania 21 Evs - 2 bonus EVs = 19

15 + 18 + 19 = 52/438 = 11.87% of population

3500 interviews per week X .1187 = +/- 415 interviews in these states => +/- 4.9% spread over 3 states.

What exactly does this tell us other than having another meaningless number to talk about...?