Morning Jay: Generic Ballot, NRCC, DCCC, and Catch a Wave!
6:30 AM, Sep 21, 2010 • By JAY COST
1. A Note on the Latest Generic Ballot Numbers. Gallup just released its latest generic ballot number, still of registered voters, and finds the Democrats up one over the Republicans. Generally speaking, the RealClearPolitics generic ballot average has shown some tightening in the last week. Is this real movement?
For starters, Gallup is bouncy. That’s just the way it is. If you look at the other polls in the RCP average right now, you find Rasmussen and Fox basically where they were in their last polls and the AP/GfK poll showing Republican improvement relative to its last poll. In addition, you have the CBS News/New York Times and Politico/GWU/Battleground Poll, neither of which has been in the field recently, so there is no way to tell whether their relatively tight numbers are a trend or just house effects.
Then you have the PPP poll, which deserves special comment. Last month, PPP switched from its “Voter” model to a “Likely Voter” model for its polls of specific races. The “Voter” model sampled Americans who voted in the previous couple elections, which meant that relative to the 2010 midterm, it was probably over-sampling Democrats to some degree. It’s good that PPP switched to a “Likely Voter” model for its individual races, but its generic ballot number is still using the “Voter" model. Unsurprisingly, it finds the Democrats with a 5-point partisan edge over Republicans, which basically mimics the D+7 2008 electorate. Now, PPP has proven itself to be a reliable pollster, most recently nailing the Delaware Republican primary battle, but a “Voter” model such as this should be approached with great caution. It’s quite unlike the other polls in the RCP generic ballot average, and it is probably over-sampling Democrats.
2. Historical Perspective on the Generic Ballot. Let me make another point about the generic ballot numbers. They cannot be read naively. The fact that the Democrats are up one in the Gallup registered voter generic ballot strongly suggests they are not actually up by one! The generic ballot has a long history of over-sampling Democrats, occasionally by a large margin. At this point in 2004, the generic ballot average was pointing to a Democratic lead of between 2 and 3 points, and on Election Day the Republicans won the House popular vote by about 3 points. The same story is true for 2002. At this point, the CNN/USA Today/Gallup poll found the Democrats with a 5-point lead among registered voters, which was generally in line with generic ballot polls from mid-September of that year. The Republicans went on to win the House popular vote by 5 points. In 1994, the Gallup generic ballot bounced around in the summer between a tie and a Republican lead, just as it has this summer, and the GOP won the House popular vote by 7 points.
Much of this has to do with the bias toward Democrats in registered voter models. Per Gallup's Frank Newport: