Final Poll: Obama 55%, McCain 44%, Other 1%
Gallup: Party Weights 39% Dem, 31% Independent, 29% Republican
Obama pary share: 91% dem, 9% rep, 48% indep
McCain party share: 91% rep, 9% dem, 43% indep
Undecided party share: 9% indep
Obama Support: 91% Democrat(91% x 39% party weight =35.49%), 48% Independent (48% x 31% Independent weight= 14.88%), 9% Republican (9% x 29% party weight=2.61%
35.49% + 14.88% + 2.61% = 52.98%...2.02% short of the 55% Gallup gave Obama. Where did they get the other 2.02%? Must have come from the 9% left of independents.
McCain Support: 91% Republican (91% x 29% party weight =26.39%, 43% Independent (43% x 31% party weight=13.33%, 9% Democrat (9% x 39% party weight=3.51%)
26.39% + 13.33% + 3.51% = 43.23%...0.77% short of the 44% given to McCain in the final poll.
So now let’s allocate the 9% undecideds (according to Gallup) to get to the final result. As we said, Obama needs 2.02% more accounted for while McCain needs 1.16% accounted for. 2.02%/31% (indep party weight)= 6.52% indep.
McCain: 1.16%/31% (indep party weight)= 3.75%
6.52% + 3.75% adds up to 10.27%, more than the 9% we said was unallocated. Gallup must have rounded the final results. We’ll just say Gallup gave Obama 6% to McCain’s 3% of the remaining independents. So they basically gave Obama a 2-1 break of the remaining undecided independents while he only got 48% of the decided vote. It’s much more likely that the undecideds break 2-1 for MCCAIN. That’s outrageous point number 1 we’ll make in this post.
Ok, bear with me now. Final Gallup breakdown is Obama: 91% dem, 9% rep, 54% indep (48+6 undecided allocation)
McCain broke down as follows: 91% rep, 9% dem, 46% indep. (43+3 undecided allocation)
There is a statistical noise of 1% for Barr and Nader, etc. Let’s just assume, for the sake of easy calculations that you can subtract 1% from any final result Obama gets in my model.
Now, let’s reweight to 38% dem, 34% rep, 28% indep just for the sake of argument. You can run your own allocations that you like better. Let’s also give McCain undecideds 2-1, resulting in 51%-49% independent advantage for Obama.
Obama: 91% dem (91% x 38% party weight)= 34.58%, 9% rep (9% x 34% party weight) = 3.06%, 51% ind (51% x 28% party weight) = 14.28%
34.58% + 3.06% +14.28% = 51.92%
McCain: 91% rep (91% x 34% party weight)= 30.94%, 9% dem (9% x 38% party weight) = 3.42%, 49% indep (49% x 28% party weight) = 13.72%
30.94% + 3.42% + 13.72% = 48.08%
So under this model we have Obama 51.92%-48.08%. We have just explained away a 4% increase for McCain and 3% decline for Obama by re-weighting the party IDs and reallocating undecided independents from 2-1 for Obama to 2-1 for McCain (which seems much more likely).
If undecided independents break for McCain 3-1 you get this result: Obama 51.71%-48.29%.
What I would be extremely interested to know is how Gallup allocated ALL undecideds. I was able to extract by calculation the approximate break down of undecided independents, and it shows a 2-1 break for Obama. I find that extremely unlikely. What I cannot calculate from the available information is how ALL undecideds broke. If Gallup gave Obama a 2-1 break of ALL undecideds then that further skews the poll. If that’s the case then you could easily turn that around to a 2-1 break for McCain and then this is at worse a dead heat if not a McCain lead using a more realistic party ID breakdown.
Conclusion and what we already knew anyway, Gallup skewed this poll hugely in Obama’s favor and it has just been debunked as unscientific.
I noticed a slight error in the last calculation, but the was skewed slightly in Obama’s favor. I showed Gallup using 2-1 breakdown for undecided independents when I should have used a 2.5/1 skewing. So, Gallup was even more in favor of Obama than I originally calculated, but the re-weighting model is still correct.
Good post.
I would only add that Obama is NOT getting 91% of the Democrat vote. So I wonder what other weighting factors Gallup uses to get the above result?
What other weighting variables would favor Obama Democrats vs. Hillary Democrats?
Rural vs. City, Black vs. White and Young vs. Old are three variables I think would differentiate an Obama Democrat vs. a Hillary (potential crossover) Democrat.
I want to see Gallup's weighting for these three variables. Or better yet Gallup, don't waste my time with your worthless poll.