With final results of the Iowa caucus as follows:
Barack Obama 37.5%
John Edwards 29.8%
Hillary Clinton 29.4%
we should revisit the question of which polling company should be trusted or not.
The winners are Insider Advantage, Zogby and DMR.
The big loser is ARG.
Much, much more analysis after the fold:
Last polls before the Iowa Caucus in order of polling dates:
Rasmussen not included because they didnt poll after christmas.
When provided, numbers including second choice reallocation are used.
Win = caucus winner projection
123 = order of candidates
RM1-2 = deviation from the relative margin between 1st and 2nd
RM2-3 = deviation from the relative margin between 2nd and 3rd
# = absolute deviation from final numbers
Ø = median deviation
MAX = maximum deviation for a candidate
The winner in the respective category is bolded!
Research 2000 (12/26-27)
Obama 29 (-8.5)
Edwards 29 (-0.8)
Clinton 28 (-1.4)
Win: wrong / 123: wrong (1) / RM1-2: - 7.7 / RM2-3: -0.6 / # 10.7% / Ø 3.6% / MAX 8.5%
Mason Dixon/MSNBC (12/26-28) (with second choice forced)
Obama 26 (-11.5)
Edwards 33 (+3.2)
Clinton 26 (-3.4)
Win: wrong / 123: wrong (1) / RM1-2: - 14.1 / RM2-3: -6.6 / # 18.1% / Ø 6.0% / MAX 11.5%
Bloomberg/LA Times
Obama 22 (-15.5)
Edwards 25 (-4.8)
Clinton 31 (+1.6)
Win: wrong / 123: wrong (2) / RM1-2: - 10.7 / RM2-3: -6.4 / # 21.9% / Ø 7.3% / MAX 15.5%
CNN/Opinion dynamics (12/26-30)
Obama 31 (-6.5)
Edwards 22 (-7.8)
Clinton 33 (+3.6)
Win: wrong / 123: wrong (2) / RM1-2: - 1.3 / RM2-3: -11.4 / # 17.9% / Ø 6.0% / MAX 7.8%
DMR (12/27-30)
Obama 32 (-5.5)
Edwards 24 (-5.8)
Clinton 25 (-4.4)
Win: right / 123: wrong (2) / RM1-2: - 0.3 / RM2-3: -1.4 / # 15.7% / Ø 5.2% / MAX 5.8%
Strategic Vision (12/28-30)
Obama 32 (-5.5)
Edwards 29 (-0.8)
Clinton 27 (-2.4)
Win: right / 123: RIGHT / RM1-2: - 4.7 / RM2-3: -1.6 / # 8.7% / Ø 2.9% / MAX 5.5%
ARG (12/31-1/2)
Obama 25 (-12.5)
Edwards 21 (-8.8)
Clinton 34 (+4.6)
Win: wrong / 123: WRONG (3) / RM1-2: - 17.1 / RM2-3: -13.4 / # 25.9% / Ø 8.6% / MAX 15.5%
Insider Advantage (1/2) (with second choice forced)
Obama 34 (-3.5)
Edwards 33 (+3.2)
Clinton 32 (+2.6)
Win: right / 123: RIGHT / RM1-2: - 6,5 / RM2-3: -0.6 / # 9.3% / Ø 3.1% / MAX 3.5%
Zogby (1/2)
Obama 31 (-6.5)
Edwards 27 (-2.8)
Clinton 24 (-5.4)
Win: right / 123: RIGHT / RM1-2: - 3.7 / RM2-3: -2.6 / # 14.7% / Ø 4.9% / MAX 6.5%
4 out of 9 polls predicted Obama as the winner, with a clear trend of later polls doing better on that score.
Only three pollsters got the final order of candidates right: Zogby and Insider Advantage! The DMR was close, but a close miss is a miss too.
Winning in the most categories interestingly are Insider Advantage and Strategic Vision. They had the lowest deviation (absolutely, relatively and the lowest maximum deviation).
Close on their heels in respect to the relative margins between 1st, 2nd and 3rd is DMR.
Zogby aint bad in absolute numbers too, pretty close to the best pollster. After all their telephone polls can be trusted!
Summarizing: Two republican polling outfits proved to have the best projection overall. So dont come up with the old tired argument, that they cant be trusted because they are a republican outfit.
On the other side, ARG has proven again, that they cant be relied upon as a polling outfit. Retrospectively one has to question their polling methodology overall. They always had Clinton in front, when nobody else did. And they were probably wrong all through the year.
The DMR proved to be a good poll, but not the best of the pack. Speaking of the Gold standard is and always has been an exaggeration, which I hope I have shown with this diary.