Pundit Fight: Cost Versus Silver

Over at The New York Times, political polling wizard Nate Silver argues that the generic ballot may be overstating Republican gains. At The Weekly Standard, political polling wizard Jay Cost says that there’s no real evidence that the generic ballot is really overstating Republican gains.

Cost vs. Silver can’t be any worse than this movie…

The Generic Ballot 101

But first, a little background on what the generic ballot means. The generic ballot question is when a pollster asks whether a voter would prefer a generic Democrat or a generic Republican. This measure tends to track the share of votes between the parties—and that share of votes between the parties can help predict how many seats a party will win in a given election. Now, its true that real campaigns aren’t between Generic Republican and Generic Democrat—but nevertheless the generic ballot still correlates quite nicely to the overall vote share.

But there’s a quirk to the generic ballot—it tends to understate Republican performance. There’s some debate about just how much, although the consensus is somewhere between 2 and 3%. That means that Republicans will traditionally get 2-3% more of the actual vote share than they get in the national vote share. That’s why the last Gallup poll in 1994 showed the generic ballot a dead heat, but the GOP went on to take 54 House seats and 7 Senate seats, retaking Congress.

Let’s Get Ready To Rumble!

But Nate Silver argues that isn’t true this year. He argues that the generic ballot is overrepresenting Republican gains instead of underrepresenting them. He looks at a series of polls conducted by the American Action Forum, a conservative group that did several polls in key House races across the country. What was interesting about the AAF polls is that they asked both the generic ballot question, and then named the candidates. The result was that the Generic Republican led by a larger margin than the actual Republican candidate.

But, says Jay Cost, there’s a reason for that—the AAF poll has a higher percentage of people who responded “Depends” to the generic ballot question than in other national polls. Which means that it’s not directly comparable with other generic ballot polls. Not only that, but in many cases the Democratic candidate was better known than the Republican challenger—or the Republican challenger hadn’t even been formally picked yet. Naturally, someone who is better known will poll better than someone who is less known. (Unless the better known candidate is really hated by the electorate.)

Of course, I’m guessing that Cost is right. Not only because I’m horribly biased and want to see Nancy Pelosi return to San Francisco where she can become a full-time Lon Chaney impersonator. But also because one set of polls has a lot less weight than 50 years of collected polling data.

But that doesn’t necessarily mean that Silver is wrong. It could be that the generic ballot is overstating Republican gains. There could be some heretofore unexplained phenomenon that’s skewing the generic ballot and causing bloggers to use the word “heretofore.” After all, the generic ballot figures have been all over the map this election cycle. One week the Republicans are up 10, the next it’s tied. The clear trend is that the Republicans are up, but by how much isn’t entirely clear.

So, we won’t really know who is right until after Election Day. And if it’s not clear by then, we can also settle it with a no-holds-barred cage match.