Last night Steven Taylor Alex Knapp posted the results of a Courier-Journal/WHAS11 Bluegrass Poll which purported to show that the Kentucky Senate race between Rand Paul and Jack Conway had tightened from a 15 point Paul lead to a statistical tie. As the comments to that post have noted, there have been several issues raised about the poll in the hours since it came out, and Allahpundit at Hot Air points out what may be the biggest problem, a bad sample:
The Survey USA poll that showed Paul up by 15 had a D/I/R sample of 47/10/42, which is in line with the 2008 Kentucky presidential exit poll sample of 47/15/38. Like I said up top, it stands to reason that Republican turnout will be better this year than it was two years ago. So what’s the sample in this new poll showing the race now within the margin of error? Why, it’s … 51/12/38. Once again, to believe Paul is ahead by only two, you have to believe that Democrats in Kentucky are going to turn out for Jack Conway in greater proportions than they did for Barack Obama. Ain’t happening, although it probably is true that the race has tightened since the poll that showed Paul up 15. Based on the spread here, figure it’s probably more in the neighborhood of eight points. Not a prohibitive lead, but comfortable.
Stuart Rothenberg, meanwhile, suggests that the initial poll showing Paul up by 15 was likely flawed as well:
When asked about the new Bluegrass Poll, which showed Republican Rand Paul’s lead shrinking from 15 points in an early September survey to a mere two points in a new survey, SurveyUSA pollster Jay Leve commented, “Whether that is a result of genuine traction for (Conway), second thoughts about his opponent, or a newly raised consciousness among voters who a month ago were not focused on the contest, I am not sure.”
Let me suggest a different interpretation: The first poll was way off. The race hasn’t closed all that much.
I’ve written often about the number of polls out there that don’t reflect reality, but it’s worth noting that news organizations that commission surveys never acknowledge that their data could be far from an accurate reflection of public opinion. That’s not surprising, of course, since doing so would discredit their own work.
But the uncertain nature of polling is one reason why my newsletter will never give its name to a survey. I don’t want to be in the position of having to defend data that seems intuitively ridiculous.
I suppose some people will believe that Paul held a 55 percent to 40 percent lead three weeks ago and that the race has closed to a razor-thin 49 percent to 47 percent now. But most veteran political observers will regard that as silly. Opinion doesn’t move that dramatically, especially since both Paul and Conway have been Senate nominees for months.
Rothenberg doesn’t address the sampling issue in the new poll, but it’s a fair point. The idea that Democrats in Kentucky are more enthused about voting for Jack Conway in 2010 than they were about voting for Barack Obama in 2008 just doesn’t make sense. As poll after poll has shown since the beginning of the summer, Republicans are far more enthusiastic, and far more likely to vote in November, than Democrats. While those numbers are likely to vary from state to state and race to race, a poll sample like the one in Kentucky, which suggests that Conway supporters are more enthusiastic to vote than Paul supporters, simply isn’t believable.
Let’s see a few more polls on this race before jumping to any conclusions. My guess is that we’ll find that the Bluegrass Poll is an outlier.
