However, if you are an investor in InTrade or other prediction markets, we will offer a word of caution for evaluating various new polling information after giving the run down of today's data. We'll also have more to say on this over the next couple of days.
On the House side, Republicans remain in a solid position to take control. The House model produced an estimated range of 234 - 236 Republican House seats or a gain of between 57 and 59 seats above their standing after the 2008 elections.
As our graph of projected Republican House gains illustrates, House Republicans have continued their march forward during a period when Senate Republican prospects have taken a step backward.
Today's charts from the micro model's individual outcome and cumulative outcomes distributions are posted below, followed by our comments/cautions on reviewing polling data.
The likely voter model that best projects this year's electorate is the likely voter model whose associated poll will be declared "most accurate" by many pundits. To be fair, there is a great deal of science that goes into developing these models, in some cases, even "rocket science." But at the end of the day, the model that worked best one election cycle may perform poorly in the next cycle.
And changes in a particular pollster's partisan makeup from one poll to the next can produce some interesting results. These comments are not meant to disparage a particular pollster, but to illustrate a point.
On September 21, Public Policy Polling (PPP) released a West Virginia survey that showed Republican John Raese in the lead by 3 points over West Virginia Governor Joe Manchin. This was followed by the release of a new poll on October 12 that showed Manchin with a 3 point lead -- a six point swing. Both polls were showed the candidates within the "margin of error."
What is interesting about these polls is that the change in the partisan makeup of the likely voters drives the change. Has the electorate changed that much in less than a month? Possibly. No one can say for certain. The point is that the two polls differ in the headline numbers primarily because of partisan make up of likely voters.
As the table below shows, if the partisan makeup of likley voters in the October 12 release was the same as in the September 21 release, the October 12 survey would have yielded a topline result showing Republican Raese with the same 3 point lead as the earlier survey.
Democrats were +22 in the October survey compared to +14 in the September survey. While backing into the topline in the September survey yields the same topline result, doing so in the October survey yields a 1 point Manchin advantage rather than a 3 point advantage. The reason for this discrepancy is most likely due to rounding.
Nothing necessarily sinister here, but looking at projected partisan makeup of the electorate may explain as much about changes in topline results as any other factor.
If the partisan makeup in the October survey was the same as in September, Manchin would have trailed 45-48 not led 48-45. As we noted, nothing necessarily sinister, but worth watching.
