Polling has two equally important parts: The collecting of people's opinions and the weighting of those opinions.
If, for example, 1,000 people were polled and 500 were conservatives, 300 were moderates, and 200 were liberals, the proportion of conservatives, moderates and liberals would have to be adjusted to accurately simulate the proportions of likely voters in the upcoming election.
And that is where the rubber hits the road.
Last spring, I repeatedly chastised pollster Charles Franklin for deviating far outside the recent history of Wisconsin voter turn-out models and weighting the sample in his poll to heavy on conservatives and too light on moderates and liberals. When the recall election rolled around, Franklin turned out to be right: Conservatives in the state turned out in unprecedented high numbers and moderates and liberals outside of Milwaukee and Madison had very poor turn-out.
Once again, Franklin is all over the news for his finding that in the span of only one month, Obama has jumped from a three point edge to a dominating 14 point lead.
Can this massive jump be believed? To answer that question, we have to look at how both the August poll result and September result were calculated.
Franklin's September poll has the Liberal/Moderate/Conservative proportions as 26/32/38. As you can see in the graph below, this is heavier on liberals and lighter on conservatives than the August Marquette poll and the earlier Recall poll.
[img_assist|nid=279615|title=|desc=|link=none|align=left|width=540|height=348]
So, a great deal of the one-month jump was because the change in the weighting, but there was also a huge, across the board movement toward Obama: He gained two points among conservatives, 19 points among moderates and a ten points among liberals.
Once again, the accuracy of Franklin's poll and other polls will come down to how close they get the weighting right.
On the one hand, if we look at the average turnout of 2006, 2008 and 2010, the breakdown of liberals, moderates and conservatives was 23, 47 and 31-- meaning this poll sample is heavier on conservatives and liberals and lighter on moderates when compared to the last three major elections and that Obama may actually be doing better in Wisconsin than this poll is predicting. On the on the other hand, the turn-out could be more like the recall election, in which case the Obama's support in Wisconsin is registering to high in this poll.
All in all, I think the turnout will be closer to the average of the last three elections than the recall election, and that Dems should be feeling good about their prospects in November.