It would take a major turn of events for the Democrats to hold onto the House of Representatives. There I said it. Why the confidence on Republicans gaining 40 seats and getting to the 218 seats needed to control the House?
The generic House ballot is tilting to Republicans in ways not seen... ever. Or as Michael Barone put it, the Republican margin currently seen is "historically unprecedented". To those unfamiliar with the generic ballot, it is the question asked on national surveys that goes something like this "If the elections for Congress were being held today, which party's candidate would you vote for in your Congressional district?" I have often (in my own head) questioned the usefulness of the generic ballot because House elections are held in districts not nationally, and surveys are only getting at most a few respondents from each district for each survey. But the fact of the matter is that vast majority of research indicates (see Charles Franklin, Matthew Shugart, and the Pew Research Center among others) that if you properly use the results from the question, you can get a pretty good idea of what is going to happen.
(he has a bunch of links on his site, I'm not going to put them here)
One of the more interested studies involving the generic ballot was carried out by Joe Bafumi (my introduction to statistics professor and an overall good guy), Bob Erikson, and Chris Wlezien. They found that by "adjusting" (the words of Andrew Gelman who linked to the study) for certain conditions, you can predict the national House vote pretty accurately as far as 300 days out. The basic findings for midterm elections are that Democrats tend to do better on the generic ballot than they do in the actual election (most studies agree on this fact), and that the party out of the White House (Republicans this year) does better than the generic ballot indicates. Both of these discoveries are good signs for the Republican party, and the signs are even better for Republicans when you look at the numbers in detail.
Yay! So how big? The numbers might stun you.
The current Pollster.com (in full disclosure, I'll be an intern at Pollster.com in the spring and summer of 2010) trend estimate has the Republicans with a 1.7% lead. If you allocate undecideds based on how those who already registered a preference say they are going to vote [Republican % of vote / (Republican % of vote + Democratic % of vote)], it's a 2 point Republican advantage (see right). For being this far out in an election (254 days), this lead is the highest since 1946 (when the generic ballot was first implemented by Gallup). What about if we take out all the polls that some Democrats would say are biased against Democrats (Rasmussen and all Republican pollsters)? The Democrats regain a lead of 1%, which readjusted is a 1.2% lead. Folks that number is still the most favorable polling result for the Republicans with . a Democratic President since 1946. Of course, many of the earlier surveys in Bafumi, Erickson, and Wlezien's data set are based off national adult samples from Gallup. So what does the latest Gallup survey of national adults (conducted earlier this month) predict? A 45%-45% tie (readjusted to a 50%-50% tie), which is is also the worst number ever for Democrats this far out with a Democrat in the White House.
How accurate are these estimates? Well, Bafumi et al.'s regression (which takes into account the generic ballot and party in the White House) predicted a 10% Democratic margin over the Republicans in the last midterm election (2006). According to data from the Federal Election Commission, the Democrats won the national House vote by 8.4%. Not a straight on prediction, but not bad either. And if anything, it overestimated the Democratic margin. Such an error should be expected with only 15 case studies (or elections), and a root-mean-squared error (a tool that measures errors in estimate and penalizes for larger errors) of 1.90%. So, these results can be trusted, but with the understanding that some error is involved. Of course with the Republicans in such a good position, the idea that this error takes away from the basic conclusion that Republicans are in a great position is not a wise one.
How will the generic ballot results from 2010 at this point translate into vote in the general election? Based off the Bafumi et al. regression (see page 6), we would expect Republicans to win the national vote by anywhere from 7.3% (all polls but Rasmussen and Republican pollsters) to 9.3% (all polls), which extends well beyond the root-mean-sqaured error. Thus, I have a hard time believing based off the polling that the Democrats will win the national party House vote.
So how will this national house vote translate into seats controlled by the Republican party in the 112th Congress? It's without a doubt a tricky question. One could imagine a scenario where the Democrats lose a majority of the national vote, but maintain a majority thanks to winning many districts by small margins and losing fewer by large margins. Yet, the evidence indicates that this is not a likely scenario. Jonathan Kastellec, Andrew Gelman, and Jamie Chandler did a nice job in a 2006 paper to show that since 1946 the margin between the percentage of the national vote and percentage seats gained by either party shrinks as you get closer to an even split of the national vote. Thus, one would expect that Republicans would win a larger number of seats relative to their popular vote as their lead in the popular vote rises.
With current polling in conjunction with Bafumi et al.'s paper predicting a Republican national vote between 53.6% and 54.7%, the Republicans could easily gain 50-60 seats from their current 178. Gains of greater than 60 seats also look quite possible. Even in the best case scenario for the Democrats, it would seem that holding the House would be very, very difficult.
It looks like a red blizzard is going to sweep through Washington in November.
I can't wait, 60+ might be low depending on how polls play out over the next 8+ months.
No comments:
Post a Comment