# Princeton Election Consortium

### Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

## On Los Angeles radio this morning

#### October 27th, 2008, 8:47am by Sam Wang

This morning I’ll be on KPFK 90.7 FM in Los Angeles. Air time: 8:20 am Pacific.

Tags: 2008 Election

### 15 Comments so far ↓

• SpxNofx

Just heard you on KPFK…very good interview. Love your election page – great work.

• Frank

In my continuing effort to convince myself that this election’s in the bag, I looked at http://election.princeton.edu/code/matlab/stateprobs.csv (8am today) for the first time. This raised a couple of concerns that I hope Sam or others will explain away.

First, Obama’s margin seems too high in some states. He’s more than 10 points ahead of Kerry (i.e., (Obama minus McCain) in the polls minus (Kerry minus Bush) in the election) in 21 states, and more than 20 points ahead in three. For example, he’s 2 points ahead of McCain in ND, which Kerry lost by 27 points; 4 points ahead in IN, which Kerry lost by 21 points; and 2.5 points ahead in NC, which Kerry lost by 12 points.

Second, although I know how the state win probabilities are estimated, in a predictive sense they seem too sensitive to deviations from ties. (I made a graph illustrating this point but can’t paste it here and will instead send it in e mail to Sam.) There are only 10 states that are not sure bets for Obama or McCain, and, aside from the tie in FL, all of these favor the leader by more than four-to-one. To continue the example of ND, IN, and NC, the probability that Obama wins all three can be calculated as 73 percent. In other words, as far as the model knows, this is close to giving the whole state to the leader.

• William

In NC, at least, about half the discrepancy can be answered for by migration. PPP’s polls showed only a 5-point gap in the vote for Bush, primarily because a lot of people in this state voted in a different state four years ago. The rest of the discrepancy can be explained by the fact that Obama’s doing 7 points better in general against Kerry.

• William

(North Carolina, by the way, is one of the fastest growing states, and in fact surpassed New Jersey as tenth-most-populous state in 2006)

• Rachel Findley

I enjoyed the broadcast. I would like to donate through the ActBlue site… but the Alaska Senate race is not on it. I am only partly acting out of rational desire to get the maximum power for my minimal dollars; there’s something sweet about the prospect of a Democratic senator from Alaska.

• Sorry, Rachel – that was an oversight. Begich v. Stevens is now listed. Thanks for catching that.

• Chris

Time for one last label on the median EV estimator?

A right-leaning group just released a Jeremiah Wright TV spot in 3 battleground states.

• Frank,

One difference in states like ND and IN is that in 2004 Kerry did not run much of a campaign in those states. Obama has been advertising in ND and IN and has made multiple visits to IN. Meanwhile, McCain has not put much effort into these states which were thought to be safe Republican states. I think this difference in campaign strategies would account for some of the polling difference between 2004 and 2008. In Republican states like OK, WY, UT, ID where Obama has not campaigned, you will see McCain still has 20+ percent margins.

• I think Mark’s points are good. Money, Charisma, Message and Circumstance have all been better for Obama. Circumstances are so good in fact, that it looks like he can overcome the percentage of racist Americans who otherwise would have decided the election.

Sam, aside from the Bradley effect, have you analyzed the racism effect? That is to say, what percentage of voters would have voted for Barack if he were white? The problem I see, from a stats perspective, is that this question is asked in so many ways it makes it difficult to find the truth, and that respondents may be likely to lie in this kind of survey.

• Dan M

I think the Alaska senate race just fell off of the knife edge. Even Alaskans can’t vote for a convicted felon. Can they?

• Dan M – good point. Note that the net overall effect of a Begich win over Stevens is that a 59-41 split is even more likely. Without Alaska, most of the remaining uncertainty is in Mississippi and Georgia.

• William

How difficult would it be to include the third-party candidate in Georgia and the probability that no candidate reaches 50%(thus triggering a runoff election)?

• I’d be surprised if there were a runoff. Third-party candidate Barr in Georgia is running in the vicinity of 1%, as are third-party candidates everywhere. Since the margin is currently larger and would go to the current leader, McCain, the current calculation seems adequate.

• William

I’m referring to the Senate election, where the third-party candidate is doing much better.

• Oh, I see. Thanks for pointing that out.

The Libertarian candidate Buckley appears reasonable by current standards – well within the range of what Republicans (and even Democrats) say they favor. I can’t tell which way his supporters would go in a runoff. It seems that they would still split in favor of Chambliss, but it’s hard to tell.

In 7 out of the last 10 polls, Chambliss has 50% or more of decided voters. I think the right approach would be to do the following: assume a 70% probability of the two-candidate margin being predictive, then split up the remaining 30% according to an informed guess about Buckley voters. I’ll consider doing this for the final estimate.