Monday, November 06, 2006

Preparing for tomorrow night 

The first Monday of any month during the school year means the department meeting, and since I also have heavy teaching that day I usually am short a post or two. Today's little different, except of course that we're on the eve of the elections.

NARN will be on the air covering the elections for AM1280 the Patriot. Most of the gang will be either at the GOP Victory party in Bloomington, but Mitch and I and a couple of researchers will be hunkered down in the Pat bunker pouring through returns in a more quiet setting. I'm expecting a late night; I saw Governor Pawlenty on his tour of the state last night and told him my expectation that I wasn't sleeping much tomorrow night. Sixty people turned out at 1030 on a nice Sunday night -- it was great and surprising to see Matt Abe from North Star Liberty and Scholar's Notebook there touring with Pawlenty.

Polls have pulled that race back to a two-point lead for Hatch; as that's your friends at the Minnesota Poll, I think that one may be over early. But the mystery is that the pullback for Pawlenty -- against whom the state DFL runs ads tying him to Bush-Cheney, thinking that's their best bet to sell Hatch hasn't moved Kennedy's numbers at all. How not? I talked to someone well connected this morning who mused about Kennedy's first win in Congress, a 155-vote squeaker in 2000. "I dreamed it would 155 again." Do I think Kennedy will win? I don't know; I will bet, however, that we're still watching those Senate returns after Pawlenty retires for the evening.

Tony -- stop reading Tradesports for these races. The markets are so thinly traded that with $20 I could have moved Pawlenty just now from 43 to 59. It's too easy to paint the tape to make those markets reliable. I'm not even sold on the predictive power of the larger races, like the contracts on Democratic control of the House or Senate. But here's what you have to understand:

Suppose two of the races for seat flips have the challenger Dem with a 90% probability of winning. That's about the odds in PA and OH. That means that the probability of both those flipping is .9^2, or 81%. Suppose your next best chance for a flip is Montana, at 70%. You now have a joint probability of .7*.9^2, or 56%. That still gets you to a 52-48 Republican Senate, assuming no other changes. It doesn't take too much to see that if you start going to the leaners, you are heading towards a couple of races where the probability of a Dem pickup is under 60%, and that gets your joint probabilities to the 20-25% level -- and much worse if Cardin loses to Steele. So a 28% probability in the Senate control market for the Democrats seems to me even a little high, but not too much so.

(You'll want a look at this from Eric at, too. A 10-point lead in the polls gives a win to the leader 90% of the time. So don't count Santorum out just yet.)

In a much more combinatoric fashion you could do the same thing with the House. Yet tonight you have trouble finding many predictions north of a 25-seat pickup. If the central tendency is 18 net Democrat pickups (my eyeballing of it) the Tradesports price of 20 or so is quite low. The variation around that 18 is too much to warrant paying 80 cents to win a dollar betting the donkeys.

My guess is that it will come down to a few late races out west -- like the Washington 8th or some of the California races, where a strong Schwarzenegger lifts all boats -- and therefore it is ever more important that we keep leaked numbers to a minimum and everyone focused on getting voters to the polls. And those of us broadcasting results up late.