This is my first diary post to any blog, so please let me know if you have any suggesions on formatting, etc. Also, I feel compelled to state that I support Obama in the primaries, but the analysis below is based on numbers and does not involve any potential for skewing toward any particular candidate.
I have been reading a number of posts recently which use polling numbers to prove or disprove that one particular candidate is ahead or more electable than another, or to predict a Democatric '08 victory or landslide.
Among the polls cited have been polls released in the last couple of weeks by Survey USA (SUSA)which have surveyed Clinton vs. Giuliani in 9 states. These show that Clinton is leading in the polls in CA, KY, NM, OH, OR, VA, and WA. Giuliani is leading in AL and MO.
These polls make little sense based on other national polls that have been released.
The 2004 vote spreads along with the recent SUSA poll spreads for these states are as follows:
2004 SUSA Difference
AL (25.7) (6) +20
CA +9.9 +17 +7
KY (19.9) +5 +25
MO (7.2) (1) +6
NM (0.7) +6 +7
OH (2.1) +3 +5
OR +4.2 +10 +6
VA (8.2) +3 +11
WA +7.2 +15 +8
A (+) indicates a Democratic advantage and a () indicates a Republican advantage. The "Difference" column shows the movement to the Democratic candidate from the 2004 actual vote to the SUSA poll.
I do not have access to the Survey USA poll results from their web site because I would need to pay to access them, but kos has posted the results on this site.
So if we look at the SUSA poll results, we could conclude that there will be a massive Democratic landslide in 2008 by carrying the usual blue states and adding red states such as KY and purple states such as VA.
But, next let's weight the SUSA results based on the proportion of the total vote contributed by each state in the 2004 election (I assume here that the proportion of the 2008 vote for each state will be the same as the proportion for 2004 - this will not be the case but it should not skew the conclusions to any great extent). The weighted average swing in the SUSA polls is 8.7% to the Democrats.
If these results are extrapolated nationally, we should expext that the polls measuring the matchup between Clinton and Giuliani should show a movement of 8.7% toward the Democrats from the Kerry/Bush race of 2004.
The problem is that they do not show this.
If we look at the only two polls measuring the Clinton/Giuliani race taken in August here:
www.realclearpolitics.com
we see that the Quinnipiac poll has Clinton up by 3% and the Rasmussen poll has Giuliani up by 7%. Bush had an advantage of 2.4% in 2004. If we subtract the swing of 8.7% from the state SUSA polls from the 2.4% Bush margin over Kerry in 2004, we would expect to see a national poll with a Clinton margin over Giuliani of 6.3%. But this is not the case.
It would seem that some, if not all, of the SUSA state polls and/or the Rasmussen and Quinnipiac polls are severely flawed. The numbers just do not add up.
So I think that it very dangerous and foolhardy to read anything into the general election matchup polls at this stage of the game. I suspect that it is particularly foolhardy to put much faith in the state polls.
I'm sure that this will not stop the number crunchers here from using these poll numbers as if they were written in stone, but I see no justification in their doing so.