Election Day! Posted by: Dale Franks
on Tuesday, November 07, 2006

I have to move out of my house today, due to the start of the slab leak/mold-related destruction, so I won't be able to blog anything until this evening. I'll also have to try to drop into my polling place between moving loads of clothes, kitchen necessities and foodstuffs to the temporary housing I'll be occupying for the next month or so.

But, I wanted to drop a final, last-minute forecast on you, according to the simple probabilities based solely on the polling averages, similar to what Jay Cost did at RCP a few days ago.

Momentum has been shifting to the Republicans over the past week. Not only have the generic polls narrowed, but the individual race polls for the seats in play have narrowed as well. House polls have become increasingly problematic (read "inaccurate") in recent elections, of course. Unfortunately, they provide the only information we have, apart from anecdotal evidence, which may, or may not, be correct.

Looking at the poll averages, however, gives us the following seat breakdowns:

That gives us the baseline expected result of 213-214 Dem seats as a result of the election. That means the Dems would still be in the minority by 7 seats. To win a majority, the Dems need another 4 seats. So, they need to run the table on all the "Leans Dem" seats and take 70% of the tossup seats. They could certainly do it, because so many of the races are so iffy, but the baseline prediction doesn't imply a Dem majority. Republicans stay in the majority simply by doing as well as probability suggests, Dems need to do better than probabilities suggest to take the majority.

In the Senate, It looks the same for the Dems: Good, but not quite good enough.

That's a pickup of 3.75 seat, i.e. between 3 and 4 seats. Of course, the numbers are pretty small, so, give the Dems the Menendez victory, and three of the tossups for a 4-seat gain. They have to run the table on the tossups to pick up a Senate majority.

The bottom line is that if the Poll averages are reasonably accurate, and the probabilistic results work out according to the baseline, then the result looks like this:

HOUSE OF REPRESENTATIVES

SENATE

Those are two large assumptions, however, and both of them are questionable in terms of reliability. Still, if I were a Democrat, I'd be as nervous as Rahm Emanuel. At the end of the day, the big variable in this one is quite simple: Who is going to get their voters to the polls? In a squeaker like this, it's going to be the biggest determinant of the day...and it can be a big determinant, making a mockery of pre-election mathematics.

I just voted this morning, so here’s my exit poll. I voted for Libertarians over Republicans and Republicans over Democrats if there were no Libertarians. For Texas governor I voted for Kinky Friedman. And here’s the one Democrat I voted for, with no regrets. Also voted ’No’ on all local propositions.

The problem with the above analysis is that it assumes that all the races are independent probabilities. They are not. They all move together to a certain extent and they are independent to a certain extent, but simply multiplying the races by some chosen multipler is not likely to capture the degree to which the different races are influenced by common national political dynamics.

In sports, independence of outcomes is a decent assumption. If during this weekend’s basketball games, Duke is a 70% favorite against Maryland and UNC is a 70% favorite against NC State in two games played at the same time in different arenas, then there is probably a 51% chance that either Duke or UNC will lose this weekend.(assuming that neither outcome affects the other).

Now, compare this to the political world.

For instance, let’s assume that Republican Senatorial Candidate A in State A is favored to win with a probability of 70% and that Republican Candidate B in State B is also favored to win with a probability of 70%. Under the simple multiplicative analysis stated above, we would predict that one of these candidates would probably win and one would probably lose.(.7 X .7 =.49).

Non-math types, however, would say that both Republican candidates are likely to win, since each has a 70% chance of winning. Who is right? Well, both viewpoints might be right to a certain extent, but the stated analysis above regarding multiplying probabilities can only be correct if all of the races are essentially independent.

And isn’t independence the crux of the issue? Most years, people say that all politics is local and the independence assumption might be more valid, but in wave years like 1994, pretty much by definition, many of the outcomes are linked to some extent, which is precisely why many analysts say that early outcomes in New York, Pennsylvania and Ohio should provide greater insight into whether a wave exists or not.

Both parties got to the polls. This prevented a serious defeat from turning into a bloodbath in the House. What happened is that Independents turned against the GOP in a shocking way.

In sum, the pollsters were generally correct; the GOP apologists at RCP and the Karl Rove and Mary Matalin types have been exposed. Anyone surprised by last night’s results simply didn’t want to know the truth.