Tuesday, September 28, 2010

Are the Experts Grossly Underestimating the Republican Tide?

In a nutshell, the answer appears to be yes, but allow me to explain.  When the collective punditry looks at an upcoming election, they have two resources: polls and election forecasting services like the Cook Political Report. (A third should be prediction markets like Intrade, but not many pay as much attention to those as I do.)

Let's examine both of these information sources, polls first. All polls sample either "registered" voters or "likely" voters, but there's a big difference between the two, particularly in midterm elections. The average turnout for the last five midterms has been 37%, and it hasn't been above 40 since 1970. Generally, pollsters only sample registered voters when elections are far off, but not always. For instance, there was a Siena registered voter poll last week in New York that showed Andrew Cuomo up by 33 points. At this point, that's number is utterly meaningless for the simple reason that Democrats are less reliable voters than Republicans. A Quinnipiac poll of likely voters done at the same time showed Cuomo only up by 6.

So, then, how accurate are polls of "likely" voters? First, you need to know how they're done. Voters are generally asked if they've voted in the last couple of elections and if they intend to vote in this one. There are some problems with this. First, it assumes someone is telling the truth. No doubt many are embarrassed to admit they haven't voted in the past and just lie about it. And who says ahead of time they're not planning on voting in an election that hasn't happened yet?

Obama was elected by a wave of voters who were, in many cases, people for whom the inside of a voting booth was a new experience. It took what they viewed as a highly inspirational candidate and a high degree of social excitement to get them there. Many of these voters are either not engaged right now, or are fully dispirited, and will likely not show up on November 2nd. And yet, they could truthfully answer affirmatively as to whether they had voted recently. They might also reflexively answer yes as to their intentions on election day, but then find themselves far too busy when the day comes.

On the other hand, many Republicans were utterly exhausted by the end of Bush's years, and just as many were uninspired by McCain. They too found themselves too busy to bother voting, and therefore would be dismissed by pollsters, per their models, as unlikely voters this time. Wrong! These are precisely the demographic that will turn out in droves. They are counting the days until November 2nd.  1994, the last Republican wave, doesn't hold a candle to how fired up people are this time.

My view is that any Republican within 5 points in likely voter poll near election day will win.

Now let's talk about Charlie Cook and his highly respected Cook Political Report (http://www.cookpolitical.com/). Right now, he sees the Republicans picking up about 40 House seats, or one more than they need to assume control. But when you really drill down, this seems quite low. There are many districts that Cook rates a "toss up" that just ain't. Pennsylvania's 8th is a good example. The latest poll has the Republican up by 14 points. How is that a toss-up? The Democrat, Pat Murphy, only gets 35%, and he's an incumbent!

New Hampshire's 1st district is another toss-up, but the latest polls there have the Republican challenger up by ten points. That means, if my theory on turnout is correct, the Republican will win by 15.
Or take my home district, New York's 19th. John Hall, the incumbent Democrat, is the former lead guitarist of the band Orleans. Remember You're Still the One? - great song. He's hirsute, moist looking guy in the middle. (This picture has appeared in my letters before - I just can't help myself.)

The 19th is, by registration, marginally Republican, and yet Hall has voted for every new Obama/Pelosi initiative. This has put him in a deeply troubled position against a formidable opponent, a Princeton-educated ophthalmologist named Nan Hayworth. The latest poll has Hall up 44-42, but let me clue you in to a little Polling 101: any incumbent below 50% near election day is in deep trouble, even if they have the lead. It means that most voters have decided to vote for the other guy, they just don't know who he (in this case she) is yet. In the 19th, Hayworth's name recognition is only at 50%, which means she has lots of upside, which Hall doesn't. Plus, if you factor in the likely turn out differential, I'd say Hayworth wins by 6-8 points, maybe more. And yet...Cook has the race as a "Lean Democrat."

(Amusing sidebar: Hall's lawn signs say "Elect John Hall." Notice why that's funny? They don't say, "re-elect.")

That's just three out of many examples, which suggests the question, what is Charlie Cook thinking? He's no dummy. To answer this, ask yourself what you would do if I asked you to spend some time building an mathematical election model. Well, your starting point would be to figure out what factors have successfully predicted election outcomes in the past. You would run correlations with various things, not unlike the process for developing quant trading models.

All right class, see if you can guess the single biggest factor, one with over a 90% correlation. Anyone? You, in the back, with pocket protector. What's that? Incumbency? Correct! You get an A.

The fact is, incumbency would have to dominate any quant election model. But this year, that's precisely the wrong factor to be leaning on in any but the most gerrymandered districts. Witness Mr. Hall's lawn signs.
My guess is that Cook relies on a combination of modeling and polls. As I hope I've demonstrated, both are highly flawed this cycle, and in fact are probably poor indicators in most wave elections. I'd say the Republicans pick up 50-60 seats this year. As for the Senate, here are the latest polls in Democrat seats that the Republicans hope to win:

Republicans are currently up in nine of these races. If you factor in my five-point rule they are up in twelve and tied in one.  Since these races are likely to be highly correlated, things still look quite good for the GOP, even if Delaware has slipped from one end of the spectrum to the other. They need a net pick-up of ten. (Note: there are no seats currently held by the GOP that appear to be in danger.)

For what it's worth, I believe we may see a nice equity market rally between now and November while these realities dawn on people. I'd probably then sell on election day.

No comments:

Post a Comment