Tuesday, October 12, 2004

My name is Robert W and I'm a pollaholic.

If you are too, you must go to Taegan Goddard
as thousands do each day. From him I learned about a great site for raw state level data 2.004.com. Also there are two cool sites which simulate outcomes.

Andrea Moro
gets Bush wins with probability 55% (incredibly close to the Iowa political market odds)

Sam Wang
Gets Kerry wins with probability 46% Again just like the Iowa political market !
He also gets a dread 4% chance of a tie in the electoral college. That would mean going to the House which votes by states one state one vote so Bush would win.


The idea is to take for each state a poll (or average of polls) and assume that the true vote Bush-Kerry will be the poll's vote Bush-Kerry plus a random variable drawn from the poll's stated distribution of sampling error (that is a normal variable with standard error almsot exactly equal to one over the square root of the sample size). Then simulate thousands of "elections"and count who wins.

I think the true chance of a Kerry win is higher given

1) undecideds split for the challenger (the rule of thumb is 60% for the challenger not 50%)
2) There are many newly registered voters in Democratic areas of Florida and the critical Ohio. This election may depend on Cleveland (a longggg way from Palm Beach).
3) Pro Democrat get out the vote drives seem very strong. Especially ACT (donate here)

Of course capturing Osama Bin Laden (unlikely) or a major terrorist attack (more likely) could change everything.

No comments:

Post a Comment