Monday, October 18, 2004
Dead even
As a public service, here are two excellent sites for keeping tabs on the election.
The first is more technical (it provides a met-analysis of state polls, a public good I've been hoping someone would provide!). It's from a neuroscientist at Princeton, Sam Wong.
The second is graphically elegant, contains lots of info, and is updated frequently, though most of the day to day variation in this site has to do with the quirky way its maintainer aggregates state polls (he doesn't). But it's nice to check every day, if you're as nervous as I am.
If you're wondering what to do with these results, here's my take:
1. The race is dead even, judging from the distribution of polls.
2. The key states up for grabs are still OH, FL, and WI; to a lesser extent PA, and then the smaller prizes of NH, NM, and NV.
3. Beyond the uncertainty from random sampling error, there is the modelling error introduced by uncertainty about who will turnout in this election. It's a historic election, with intense hatred of the incumbent in some quarters, fear about war and terrorism, the memory of a close race in 2000, and massive new registration. Good luck predicting what will happen.
4. Many people, even sober social scientists and veteran Washington journalists, think there is a strong chance the election will be decided by cheating on the part of Bush, especially, but not exclusively in FL. More on the evidence of this later.
5. You may have seen discussion of ties in the polls going to the challenger, because people undecided at this point aren't going to go with the incumbent. This LAT story reviews the evidence. I think this one is so close, and the race so unusually heated, that I'd be reluctant to expect it to run true to form. But if it helps you sleep at night, read the article.
The first is more technical (it provides a met-analysis of state polls, a public good I've been hoping someone would provide!). It's from a neuroscientist at Princeton, Sam Wong.
The second is graphically elegant, contains lots of info, and is updated frequently, though most of the day to day variation in this site has to do with the quirky way its maintainer aggregates state polls (he doesn't). But it's nice to check every day, if you're as nervous as I am.
If you're wondering what to do with these results, here's my take:
1. The race is dead even, judging from the distribution of polls.
2. The key states up for grabs are still OH, FL, and WI; to a lesser extent PA, and then the smaller prizes of NH, NM, and NV.
3. Beyond the uncertainty from random sampling error, there is the modelling error introduced by uncertainty about who will turnout in this election. It's a historic election, with intense hatred of the incumbent in some quarters, fear about war and terrorism, the memory of a close race in 2000, and massive new registration. Good luck predicting what will happen.
4. Many people, even sober social scientists and veteran Washington journalists, think there is a strong chance the election will be decided by cheating on the part of Bush, especially, but not exclusively in FL. More on the evidence of this later.
5. You may have seen discussion of ties in the polls going to the challenger, because people undecided at this point aren't going to go with the incumbent. This LAT story reviews the evidence. I think this one is so close, and the race so unusually heated, that I'd be reluctant to expect it to run true to form. But if it helps you sleep at night, read the article.