Why Republicans win 70 seats; an explanation of my projection

Jon Ralston:

First, let me be clear on this tradition of predictions. It is not a wish list but a walking out on a limb, so I can either crow afterward or eat same. I base them on data I am privy to and my gut. I have had much success in the past -- look it up. But if ever there were a year for my lifetime batting average to take a hit, this is the one.
Every word of this rings true for me. Of course, I say something like this every election. But it's really true this time.

This election is truly difficult to predict. On the House side -- as this post deals with my projection that the Republicans pick up 70 seats -- most of the seats that were originally expected to be the battleground appear to be leaning towards Republicans at the very least, even if you're a wave-skeptic. The seats that are now the true tossups usually have less than first-string Republican candidates and a complete dearth of independent public polls.

These questions aren't reflective of the model by which the projection was made, but are certainly indicative of how I view this election.

1. What is the most analogous cycle? 2008? 2006? 2002? 1994?
A pretty key question that will determine how you project the election. I'd argue that the view from 10,000 feet looks similar to 1994, and arguably much worse for Democrats. There are simply way more seats in play this election, mainly due to two factors: 1) Democratic gains, and 2) redistricting. The first, Democratic gains, is fairly obvious. Democrats won lots of previously Republican held seats in the 2006 and 2008 elections. These are swing districts with non-entrenched candidates who have elected Republicans in recent memory, moreso than in 2006 or 1994. The second is redistricting: we're in the last election of the 10-year cycle that occurs between the Censuses. Districts change and demographics are farthest away from the original cartographical intent.

Also, I'd note that Gallup's generic ballot is at an all-time high for Republicans. The high turnout model has a 52-43 Republican edge, while the low-turnout model has Republicans at 55-41, whereas in 1994 it was 53.5 R - 46.5 D. Big difference -- this year is twice the 1994 gap, and Gallup's polls have historically been very close.

2. What will turnout be like?
Your answer to the first question will certainly affect your view of turnout.

And wait! I referred to the low turnout model as if it was an apples-to-apples comparison to Gallup's historical results. But, aha! It is the same model! The low-turnout model -- the one most favorable for Republicans -- is their traditional model. It caps turnout at 40%; basically choosing the 40% of voters most likely to vote and throwing away the rest of the sample. Crazy huh? Well, not only has that been predictive for a long time, but we haven't had over 40% turnout in a mid-term election for 40 years. Gallup's high turnout model takes the top 55%, which is more reflective of a presidential campaign. So if you think Obama can turn out his voters like it was 2008 again, then you might think Gallup's high-turnout model is worth using.

Of course, even Gallup's high turnout model has Republicans leading 52-43...which is higher than Republicans have ever polled in Gallup before an election, even 1994.

3. Is Democratic enthusiasm spiking?
While people quickly forget, Republican enthusiasm spiked right before the 2006 election. Partisanship finally kicked in a little bit...and that means that Republicans only had a 59 seat swing. I have no doubt that Democratic enthusiasm is in fact spiking right now...it's just highly unlikely to spike enough.

Generally speaking, I'm not a fan of anecdotal evidence, so of course in this circumstance I'm just going to cite an anecdote: Obama's last rally in Cleveland today was half-full. 2008, this is not, even though Nate Silver still makes assumptions like it is.

Frankly, having gone through a fair amount of crosstabs in the last few days, it appears to me that pollsters are being overly influenced by the 2006 and 2008 election results. Now, perhaps those elections really were a paradigm shift; but it's at least as likely that those results are unduly influencing partisan turnout models.

4. What does a 46-44 poll mean?
It's a tossup, right? Wrong. I'd argue that a Democratic incumbent who is polling at 44% and down 2 is a likely loss given the enthusiasm gap.

Finally, I leave you with a Rorschach test: TX 27. Ortiz v Farenthold. Obama won the district in 2008 53-46, which is exactly the national result, if I recall correctly. However, W beat Kerry 55-45 as well as essentially tied Gore 50-50. Ortiz carried the district by 20% in 08 against a perennial token opponent who spent very little money. The only poll we have is by a Republican polling firm that has Farenthold up 8. Silver has Farenthold as a 23% chance to win. I have Farenthold as about a 40% chance. However, looking through the rest of OnMessage's polls for the cycle, I have to conclude that their numbers appear to be generally speaking within an acceptable range in other races. Congressional races are not independent events.

One thing I'd note: at the time I predicted 70 seats, Larry Sabato had the GOP gain at 55, Nate Silver at something like 51, Rothenberg says 55-65, Charlie Cook has the number in the 50s. Point being that I could easily pick a lower number (eg, 61) and lower my chances of looking silly, while still looking better than all the professional political pundits. The smartest thing for me to do, from a reputation standpoint, is just to pick 61 and say I got it better than everyone else.

On the other hand, I was upset when the Texans picked Mario Williams. I screwed that up; mea culpa.

Posted by Evan @ 10/31/10 08:40 PM


Previous Entry | Home | Next Entry


No comments yet

Add Comments

No flames or impolite behavior. HTML will be stripped. URLs will be transformed into hyperlinks.