Prognosticators: Watch out for candidate polls!!

[longship]

Pollsters use models and scaling on their raw results. This supposedly aligns the poll data with the reality of the district or state to which the poll allegedly applies.

I believe that the electoral dynamics this year are dramatically skewed by any number of elements which place this election outside the parameters used by pollsters. In particular, Zogby has been criticized for overstating Republican strength in many of their polls. Other pollsters are undoubtedly doing likewise.

This choice of models is what we're seeing in the chaotic nature of individual polls on individual races. They are attempting to get it right, but the predictive value of their models is breaking down.

In this election, any increase in usual turnout for Democrats and independents, could easily skew the results outside the poll models. Furthermore, there are going to be many Republican defections this year, if the number of embattled red district Repukes is any indication.

Additionally, we are still seeing a huge number of polls with a considerable block of undecideds. This cannot be correct at this late date. Are people lying to the pollsters? If so, what does that say about the rest of their numbers? I can't think of any reason for so many undecideds at this late date.

While you are watching the opinion polls, keep in mind that they are most valuable when used cumulatively, and to show trends over rather longer time periods than a few days. The specific values in these polls are most certainly going to be in error.

I'm looking at these things right now.

1. Races where there are still many undecideds. I think we'll see these people break hugely for the Democrats on Tuesday.

2. Candidates where the consensus has the candidate at or above 50 percent. Unless the consensus polling shows a very close race, these are in effect a done deal since no number of undecideds are going to change the race.

3. Races which have consistent trends in one direction. If the direction of the trend is toward the Dems, I flag the race as a possible Dem pickup.

4. Where polls are flipping back and forth, or there is no clear consensus among the pollsters, I discount *all* the polls in that race as outliers. If there is consensus on undecided percentages I might pay attention to that, but reject the rest.

I still see 7-8 Senate seats, with six being fairly sure.
I still see upwards of fifty or more House seats, with about three dozen being pretty sure.
I still see Dems losing *zero* seats (other than Lieberman :-( )

The Dems are going to get all the breaks on Tuesday.