Alan Abramowitz on turning House votes into seats

Emory University political scientist Alan Abramowitz sent out a useful analysis today that he's given me permission to post:

From fivethirtyeight.com today:

[T]here is another issue at hand: how much does the generic ballot really tell us about what will happen on Election Day? It might be the case that the generic ballot is fairly stable, but that doesn't necessarily mean it's all that useful an indicator. In addition to the fact that the consensus of polls (however careful we are about calibrating it) might be off in one or the other direction, there's also the fact that the thing which the generic ballot is ostensibly trying to predict -- the national House popular vote -- is relatively irrelevant to the disposition of the chamber, or the number of seats that each party earns. Instead, what we want to know is how the generic ballot translates into each of the 435 congressional districts; this is the sort of problem that we're hard at work upon.

Nate provides a lot of excellent analysis. But there are two pretty silly statements here. First, the generic ballot is a pretty good predictor of both the national popular vote and the national seat results. Second, the national popular vote is a very good predictor of the overall seat results. It definitely is not “relatively irrelevant” to those results. For all House elections since WW II, the correlation between national vote share and national seat share is a whopping .93 (click graphic for larger version):

For more on how the national vote translates into seats in the House, see Andrew Gelman and his co-authors on the 2006 and 2008 elections.

Update 8/19 10:01 AM: Silver responds here. For more on how the generic ballot can be used to forecast House election outcomes, see Abramowitz's 2006 PS article (PDF).

[Cross-posted to Pollster.com]

Comments

[David in Cal]

Good point, Brendan. Sounds like Silver is trying to boast that what he's working on will be better than the overall average approach, and he denigrated that approach too much.

Silver's main assertion is that the generic ballot is not a good predictor of national seat results. It would be interesting to see a chart like the one above testing that assertion. That looks tricky to me. One would have to consider the generic ballot as of various points in time.

Also, past generic ballots may have used different methods, which may not be directly comparable to today's methodology. Presumably the "generic ballot" means the RCP average or something like it. Past generic ballots may not have been tallied or may have used averages of different polls.

[Fargus]

Brendan,

I think the other thing to keep in mind is that, reading Nate Silver's post, I think he conflated two things, but what he really seems to be talking about is how well the generic ballot can do in helping predict the results of the 435 individual races. It can do quite well to predict the overall disposition of the chamber, assuming the generic ballot accurately predicts the national vote share, but it's indeed quite possible that it's relatively irrelevant to the results of any particular district, other factors considered

[Jinchi]

But where is the generic ballot at today? Rasmussen has Republicans up by 8. Gallup has the generic ballot tied. And the rest of the pollsters fall everywhere in between. That seems like a huge range to me. Is this kind of scatter normal for these polls? Or is 2010 unusual.

[Matthew Huntington]

Sorry, I just don't see it.

Looking at the pdf from 2006, there's only an .83 correlation between generic vote and actual vote. And then there's only a .93 correlation between actual vote and seats taken.

That's what, a .77 correlation from generic vote to seats taken as a base? There has to be a better indicator than that.