You guys are poll obsessed! Anyway, I have a new theory: the more random flucuation seen in a given poll the more comments it will generate. But in order to test this we'll need to know which polls are stable and which are less stable. One approach is to compute the standard deviation of a tracking polls released numbers (which are typically based on 3-day rolling averages but ignore that issue for now). Here are the results for the 7 daily tracking polls over the last few weeks (which excludes Battleground):
The smaller numbers reported above represent less deviation from the pollster's mean. Many things can affect random flucuations but one obvious one is sample size. Two of the pollsters near the top of our list here are Rasmussen and Gallup which utilize the largest samples (Rasmussen the most with 3,000 interviews). What about ABC/Washington Post? Their results are based on fewer interviews (1,580 for the most recent edition) but they use a 4-day rather than a 3-day rolling average. That means each day's reported result is based on the average of the last 4 days which reduces the impact that one "weird" day will have on their results.
Zogby has the most flucuation. Surprise! Interestingly, their sample size (1,200) is actually larger than those of R2K, Hotline and TIPP. Something else may be going on. I arrived at the numbers reported above by computing the standard deviations for each pollster for Obama and McCain and then averaged those two together. Separately the results look like this:
ABC Obama: 0.64
ABC McCain: 0.72
GallupObama: 0.80
Ras Obama: 0.80
Ras McCain: 0.80
R2K Obama: 0.92
GallupMcCain: 1.02
TIPP Obama: 1.09
HotlineMcCain: 1.11
TIPP McCain: 1.25
ZobgyMcCain: 1.29
ZogbyObama: 1.36
HotlineObama 1:45
R2K McCain: 1.57
In this look, the most unusual pollster has been R2K which has shown extreme stability for Obama who has been at 51 +/- 1 point almost every single day. They have McCain, on the other hand, going from 40 to 45 over a five day period recently. The most charitable explanation for this is that maybe McCain really did move that much in less than a week's time. The problem is the other trackers didn't really show much McCain movement at all over that period. It was R2K/Kos more than any other pollster that was responsible for the "tightening" theme from last week.
Now all I have to do is record the number of Quick Hit comments generated by the trackers mentioned above and then run a simple correlation..... a project for another day.
