I felt like I should talk about this for two reasons. First, because I haven't seen anyone else mention it yet. And second, because I badmouth Zogby all the time. He is the butt of every one of my pollster jokes. Literally.
So it was a shock to me, that when I compared the election results against November polls, I found Zogby near the top.
Below I have listed Zogby's November polls with the real results in parentheses.
FL: Obama + 1.2 (O+3)
IN: McCain + 5.3 (O+0)
MO: Tie (O+0)
NC: McCain + 0.3 (O+0.4)
NV: Obama + 10.8 (O+12)
OH: Obama +2 (O+4)
PA: Obama + 9.8 (O+11)
VA: Obama + 6.4 (O+5)
WV McCain + 9.9 (M+13)
So Zogby's November polls were off by an average of 1.9 points. They all underestimated Obama's support to a greater or lesser extent (except Virgina). This is actually good, because the rule of thumb is that a good ground game will net you two points. In fact, Nate calculated in his post "The Contact Gap: Proof of the Importance of the Ground Game?" that Obama's ground game should be worth a little less than two points. Also note how he nailed Nevada. The margin there was quite a shock to me at 12 points. I think 538 was predicting a margin of 5 points.
In light of this, Zogby did a phenomenal job. If you are generous enough to throw out the Indiana poll (which was a big surprise to everyone), then he was off by an average of 1.25 points per poll. That's like predicting 49-48, when the result is actually 50-47.8. On average!
For someone like SurveryUSA, whom Nate used as the Gold Standard for this cycle:
GA: McCain + 7 (M+5)
FL: Obama + 3 (O+3)
IN: Tie (Tie)
MN: Obama + 3 (O+10)
MO: Tie (Tie)
NC: McCain +1 (O+0.4)
OH: Obama +2 (O+4)
PA: Obama +9 (O+11)
VA: Obama +4 (+5)
WA: Obama +16 (O+16).
The SurveryUSA polls had an average error of 1.6 points. As with Zogby, it usually under-counted Obama's support. You can certainly see why SurveyUSA is so highly regarded, but note that Zogby's average error of 1.9 points is almost the same.
Especaily compared to other pollsters. Rasmussen, for example, had an average error of 3.9 points on 13 polls. Ouch. Also under counting Obama. No surprise there.
My hat's off to Zogby on this one.
::
As one last note, DailyKos didn't have any polls in November, but if you look at their polls for 10/28-10/30 you find an average error of 3.5 points on 11 polls, or 3.1 points on 11 polls if you leave Alaska out (damned Steven's effect.)