Daniel at Read Meat Conservative explains how Survey USA's Kentucky Senate poll that rather conveniently was released this weekend illustrates how bogus party ID weightings can be used to distort a survey. After showing Palin-endorsed Rand Paul with a 15 point lead over Democrat Jack Conway only three weeks ago, now Survey USA purports to show Conway within two percentage points. Here's how S-USA fudged the numbers:
The poll taken earlier this month had a party ID breakdown of 47 D 10 I 42 R. That is actually slightly less favorable to the GOP than the disastrous 2008 turnout of 47 D 15 I 38 R (in which McCain still won). Remember that there is no doubt that there will be more Republicans turning out to vote in Kentucky this year. The only question is how much more? However, nobody disputes that the turnout figures will be much more favorable than those of the past two elections cycles.Daniel cites Public Policy Polling (PPP) as another textbook example of how manipulating a survey's party ID weightings can make the poll results show anything an unscrupulous pollster wants them to show. Earlier in the election cycle, PPP polling showed Sen. Burr Richard (R-NC) virtually tied with his opponent, Democratic Elaine Marshall. This is because, Daniel argues, PPP was employing a similar turnout model to the one used by Survey USA in its latest survey. Since changing its methodology to employ a a more realistic turnout model, most PPP polls are showing the true advantage Republican candidates have over Democrats in an election cycle where voters are fed up with six years of a Congress under the control of radical leftist Dems.
This brings me to the most recent Survey USA poll that shows Paul leading Conway by a paltry 49-47 margin. The party ID weightings of the survey were as follows: 51 D 12 I 38 R! Does anyone really believe that a majority of the voter turnout in a staunch conservative state, in a staunch conservative year, in a state where Obama's approval ratings are in the toilet, will be Democrats? Is it really possible for voter turnout models in 2010 to be worse for the GOP than 2008?
The most recent survey by the more reliable Rasmussen Reports shows "Burr enjoying his best showing to date, with 54% support to Marshall's 38%." Likewise, Rasmussen's poll in the Kentucky Senate race conducted earlier this month showed Rand Paul with "his highest level of support since winning his party’s primary in May" -- leading Conway handily 54% to 39%.
As Daniel says, "The devil is always in the details." That's usually where you'll find a Democrat with a heavy thumb on the scales.