Charles Franklin, who was widely criticized in 2010 for tweeking his poll results to fit the wishes of the rightwing group WPRI, has released another stinker.
The main problem with this poll is that their sample is absurdly skewed to the right.
Let me explain: At the beginning of any poll, you need the correct sample in which to extrapolate to the rest of the population. This sample should be of people that are actually going to vote and representative of the population that will vote.
In this poll, they called, at random 4,916 numbers of both cell phones and land lines. Of those 4,916 calls, 701 actually completed the survey.
Of these 701 people, 92.9% indicated they were either "absolutely certain" or "very likely" to vote, 4.6% said they were "50-50," and 2.3% said they would not vote.
Of these 701 people, 292 described themselves as "very conservative" or "conservative." Of these people, 85.6% said they were "absolutely certain" to vote and 7.9% they were "very likely" to vote. Only 6.5% said they were "50-50," "will not vote" or "don't know." In other words, this an incredibly motivated sample of conservative voters.
Another 228 people, (again out of the 701 sample), described themselves as "moderate." Of this group, 90.8% said the were either "absolutely certain" or "very likely" and only 9.2% indicated one of the less firm responses. Again, a very motivated segment of voters.
The last part of the 701 sample were 145 that described themselves as "liberal" or "very liberal." Of this group, 94.5% indicated a strong motivation to vote and 5.5% gave a less firm response.
And 36 people refused to identify their political leanings.
In other words, of the 701 people polled, 41.6% were conservative, 32.5% were moderate and only 20.7% were liberal. And, not only was the deck absurdly over-stacked with conservatives, they were realllllllly motivated conservative voters.
Who do you think an extraordinarily motivated conservative voter is going to support?
Or, look at it this way: If you had a group of ten people and four were conservatives, two were liberals, and three were moderates, what are the odds that five of the ten would support governor Walker? That, in a nutshell, is the only this absurd poll told us.
This is, of course, not what the electoral make-up of the recall election will look like.
If we average the last three exit polls in Wisconsin (2006, 2008 and 2010), we find that the actual breakdown of the Wisconsin electorate is 22.7% liberal, 46.7% moderate, and 31% conservative.
Using the weighting described above, the poll results look like this (note: 5.1% of poll respondents refused to give a political idealolgy, so an adjustment has been made for that as well):
Walker Favorability: 48.4% Unfavorability: 51.4%
Walker 46.4%, Obey 49.0% Margin = +2.6%
Walker 48.4% Barrett 50.3% Margin = +1.9%
Walker 49.4% Falk 48.5% Margin = -.9%
Walker 48.2%, Cullen 45.2% Margin = -3%
That's right: If you figure that most undecideds usually go to the challenger, ALL the Democratic candidates would beat Walker. In addition, this population still undercounts the number of union households in Wisconsin and that would add a two or three points difference to these results as well. These "in the right ballpark" weighted results are really not all that surprising considering that the same poll showed that Walker is viewed favorably by only 41.7% of moderates.
NOTE: In my initial post, I used the breakdown of 30% liberals, 40% moderates and 30% conservatives. I still beleive that this will be ultimately be the breakdown in the recall election, where I think liberals will be more motivated to vote than consevatives. However, I think it is "more prudent at this juncture" to use the exit polls.