News from the Votemaster

We have a new feature today. If you click on a state, you get not only a graph with the 2008 polls, but also one with the 2004 polls for comparison purposes. However, two footnotes are in order. First, the site went live May 24, 2004 and no data was collected prior to that date; for 2008, all polls taken this year are included. Consquently the scales are different. Second, it got pretty messy in Oct 2004. But since the idea now is to compare summer of 2008 with summer of 2004 it will do for the moment. Starting in the Fall, we will go to an expanded scale.

Steve Lombardo, a Republican pollster, wrote a column on Obama's overseas trip at pollster.com. In short, he says the trip was a brilliant idea. According to Lombardo, elections are about impressions, not about policy papers. The more-or-less continuous news coverage showing Barack Obama meeting with world leaders, talking to the troops in Iraq, and today giving a speech in Berlin to a huge crowd makes him look "presidential." Lombardo agrees with many other political experts that this is Obama's election to lose. If he can convince enough people that he is "presidential" enough (whatever that means), he wins.

This is not John McCain's week. Not only is Obama on TV, radio, and the Internet 24/7 this week, but even when McCain has a clever idea for publicity, the fates do him in. He was going to take a helicopter out to a drilling rig off the Louisiana coast to emphasize his plan for more oil drilling in the U.S. to lower gas prices, but bad weather forced him to cancel the trip. Or at least that is what he said. However, there is a huge oil spill on the Mississippi River near New Orleans. Could it be that what McCain is really worried about is ferrying hundreds of reporters to Louisiana, getting two paragraphs about his visit to the rig and then 10 paragraphs about the oil spill, reminding the voters of the dangers of drilling off the coast? The McCain campaign didn't issue a statement about the oil spill.

Rasmussen ran a poll asking people who they think John McCain will choose as his running mate. The winner? Mitt Romney with 46% of Republicans expecting this. Of course few, if any, have inside information. This kind of poll is often discussed in statistics classes usually stated as "Suppose you were to ask a million people what is the height of the emperor of Japan?" The answer would be very precise (reproducible) because the sample size is so large, with a margin of error around 0.1%. However, it would not be very accurate because none of the respondents have a clue what the height of the emperor is. This question, and also the Veep question, illustrate the difference between precision (large sample, small margin of error) and accuracy (result is close to the true value).

CQ Politics has a story giving four basic principles for Veep selection:

1. The Veep can't oppose the party on abortion.
2. Don't choose a senator who would be replaced by one from the other party.
3. Don't pick someone who might embarrass or outshine the candidate.
4. Make sure the Veep adds something to the ticket.

The first one is obvious: the base would rise up and smite the candidate. Former Pennsylvania governor Tom Ridge (R-PA) is out for McCain and so is Sen. Chuck Hagel (R-NE) for Obama. The second one precludes Sen. Jack Reed (D-RI) because the governor of Rhode Island is a Republican. A corollary says don't give the other guys a free governorship. A McCain choice of Gov. Bobby Jindal (R-LA) would make Lt. Gov. Mitch Landrieu (D-LA) governor of Louisiana (and launch him to challenge Sen. David Vitter (R-LA) in 2010. The third principle eliminates Sen. Hillary Clinton (D-NY) since she would outshine Obama with certain groups and her husband might embarass him. Gov. Sarah Palin (R-AK) and her motorcycle might distract the media from McCain himself too much. The first three principles can be summarized as "Do no harm."

That leaves #4: try to do some good. A Veep candidate could help win a specific state (e.g., Gov. Charlie Crist could help McCain in Florida), win a particular demographic group (e.g., Gov. Bill Richardson could help Obama with Latinos), or something else, but the Veep should bring something to the ticket.

We have another Florida poll today. Yesterday ARG reported that McCain is ahead by 2 points in Florida. Today Rasmussen confirms that it is a statistical tie there. His numbers put Obama 1 point ahead. Either way, Florida is going to be a huge battleground. Both sides are pouring money and staff into the state. McCain has the advantage of a popular governor working for him. Obama has the advantage of a net 90,000 new Democratic voters registered since 2004. While Florida isn't the only battleground this year, it is a biggie.

A new poll in Minnesota putting Obama ahead 49% to 37% can't help Gov. Tim Pawlenty's chances of snagging the VP nomination. Having a local boy as running mate helps a little bit, it can't overcome that kind of deficit. If McCain thinks he is going to lose Minnesota big time, what's the point of putting its governor on the ticket?

We have two new Senate polls, both surprising. In Minnesota, Al Franken seems to have caught up with Sen. Norm Coleman (R-MN), in contrast to many earlier polls showing Coleman safely ahead. Similar, in New Hampshire, Sen. John Sununu (R-NH) appears to have caught up to former governor Jeanne Shaheen (D), when polls for months show her with double-digit leads. When a poll contradicts so much other evidence, don't take it too seriously, not in Minnesota (sorry Democrats) and not in New Hampshire (sorry Republicans). Remember than 5% of the time the true value falls outside the margin of error. For the standard state polls, the MoE is about 4-5%. So there is a 5% chance that Franken's real score is outside the range 38% to 48%. Since there are well over 20 polls a week, one or two of these outliers are going to show up every week. It's normal (sorry statisticians).