I don't have any special knowledge or any special sources. I'm not privy to anything any other American can't find on the web. I'll start from the map of house races posted at Real Clear Politics. At the time of this writing they show 224 seats leaning, likely or safe for the Republican Party, 167 seats leaning, likely or safe for the Democrat Party and 44 rated "toss-up".
If we are to assume that the polling is indeed predictive the cautious prediction would be to split the toss-ups in half. That means an ultimate split of 246 R, 189 D. That would be a gain of 68 for the Republicans. That's already huge, and assuming that they do as they have promised and actually fight rather than try to make friends it's a great day.
A more historically-based prediction says that the party benefiting from a "wave" event gains 70% of seats "in play". If we call all the leaners in play as well as the toss-ups then there are 106 seats "in play" for a gain of 74. That's interesting in as much as it's 1 seat shy of the biggest gain in over a century.
My assessment is based on 3 factors:
1. Most polls this cycle are based on turn-out models which either copy or weigh heavily the 2008 turn-out. I can't seen any reason to believe that is right.
2. "TEA Party" voters are a combination of fiscally-conservative Republicans and fiscally-conservative socially-liberal Independents who voted for BHO and disgruntled conservatives who didn't vote last cycle. All three of these categories are going to turn out en masse. The latter of the three can't be accounted for by a turnout model that is based on 2008 nor 2006.
3. A very high-proportion of the toss-ups and leaners are Democrat incumbents who poll under 50%. Think about that, even using a model that clearly gives the Dems too high a turn out and in no way counts the people who sat out '08 and '06 these Dems aren't polling 50%.
For these reasons I believe that this will be a bigger loss for the Dems than most people are willing to risk predicting.
My prediction is the House will end up 258 R, 177D or a shift of +80 R. That is one seat more than a complete reversal of the '08 result (not of the shift but of the final 2008 257 D, 178 R split.) There it is. I'm on the record. If I'm wrong, then I'm wrong. Wouldn't be the first time.
Now consider this. What if these toss-ups are all mis-characterized and all 44 of them go to the right. In that case the final split would be 268 R, 167 D or +90 R. I think that's possible, but I wouldn't bet much on it.
It's become a tired phrase, but it's true: It's all about turn-out. My feeling is that, especially in light of observation 2 above, the turn-out is going to blow away many many more Dems than most people are willing to admit.
What about the Senate?
The map I'm looking at right now has 45 R, 48 D+I and 7 toss-ups. The toss-ups are: California, Colorado, Illinois, Nevada, Pennsylvania, Washington and West Virgina.
Among those there are only 2 in which I hold any doubt. So 50 R and 50 D+I seems the most likely outcome on paper. I would actually like that a lot. I would prefer for the Dems to hold the Senate rather than have Republicans with a bare majority. This is a political calculation. Contrary to popular belief, 51 seats in the Senate is not the condition of being "IN CONTROL". Not at all. There is a reason why the Dems had to steal an election in Minnesota in '08. (BTW Minnesota, congratulations on that Banana-republic election-system you've created for yourselves.) The reason is that since 2000 it's been repeatedly demonstrated that if you have less than 60 seats in the Senate, you are NOT "in control".
I would very much like to see Republicans in control of the Senate, but that is simply not possible in this election cycle. Politically it is a stronger position to not be the party that voters are going to be told every day is "in control" when you are not in fact "in control". Also, I have a lot less confidence in Senate Republicans to be confrontational, and I believe that being confrontational is required to maintain the support of the TEA Party voters through 2012... and they WILL be there in 2012.
So,
Two more issues I haven't discussed here but which I have discussed in the past.
1. Political polling is not an exact science. Not by a long shot. Despite claims of margins of error in the range of +/-3 to +/-4 for the typical opinion poll, they are a lot less reliable than that. Those margin of error claims are based on an assumption that the sampling model is completely accurate. There is in fact no way to assess the accuracy of a sampling model. I know people pretend, but it is simply an impossibility.
This is why you get one polling agency showing a candidate +7 and another showing him -3. The sampling model makes for the difference and they fundamentally cannot both be right. In this cycle the worst offender in oversampling from one side beyond any reason is the polling agency PPP. They are oversampling Dems by pot-head-fantasy quantities.
2. RCP bases its analyses on the average of several polls by different polling agencies. Statisticians should cry "foul" at this. Different agencies use different sampling models so the populations represented by these various polls are in fact quite different. Therefore it is a violation of responsible statistical inference to average the results and attribute any meaning to them.
Moreover this is also very much a matter of "Garbage In/Garbage Out". Once you average in one or two ridiculous samples (like say PPP, Zogby and Newsweek), the number you get out the back end is not merely meaningless, but in fact predictably misleading.
So when you look at some of these Senate race predictions, have a mind to whose polls they are using in their average. Then remind yourself that the average should be suspect to begin with.
No comments:
Post a Comment