Roger on John

My friend Roger really enjoys probabilities and statistical analysis. I remember one time we went to a resteraunt and there were dice on the table which he rolled and explained a lot about probability to me. I am not sure how much I retained, which is why it is a pleasure to have him post today about the Presidential race. Here are his views on Kerry's chances in November. He mentions Maine because I was curious about its' importance.

Moe,

I have just discovered the poll tracker at MSN Slate. It is the best I have seen. It combines the information from all of the polls with good judgment as to which polls can be trusted and which cannot. They have Kerry winning 276 - 262. They conclude that Kerry will win all of the Gore states except Iowa. In Maine, they expect him to win the popular vote but lose one district. Slate predicts Kerry to win Ohio. This gives him the election.

The prediction at electoral-vote.com (EV) is Kerry by 301-237. There are two major differences between EV and Slate. First, EV relies on the latest poll numbers even if they come from an organization that may be unreliable. Slate looks at all recent polls. Second, EV assumes that the undecided voters will break 2-1 in favor of Kerry. The 2-1 phenomenon seems supported by results from past presidential and non-presidential elections. However I am skeptical that it can be so readily applied in this election.

Now that you have heard all of that, let me update my own probabilistic model.

Using Zogby as the primary reference, Kerry is set to win the Gore states except for Maine where he will likely win 3 of 4 electors. Kerry is also set to carry New Hampshire. This gives Kerry 263 electoral votes. To win the election, Kerry needs an additional 7 electoral votes. There are 6 states with 7 or more electoral votes where Kerry has a chance to win. Here they are with the latest Zogby numbers:

Ohio: Bush by 3 points
Florida: Bush by 1 point
Missouri: Bush by 3 points
Virginia: Bush by 3 points
Tennessee: Bush by 2 points
North Carolina: Bush by 3 points

I assess the likelihood of Kerry winning in Ohio at 1/4, and Florida 1/3. I assess the likelihood of winning one of the remaining four states at 1/10. But note that if Kerry were to win any of these four states it would signal that Bush has crashed and burned. In that case Kerry would likely carry Ohio and Florida as well.

Combining the probabilities, the probability that Kerry wins the election is 0.55.

An interesting situation can occur with Arkansas. If Kerry were to win Arkansas while losing the other swing states, the result would be 269-269. With Maine splitting 3-1, Arkansas has no relevance to the election.

Roger

Mr. Kerry you wouldn't believe the lobster you can get in Maine - they're delicious! Come on out and touch a hand and make a friend if you can.