Election Day

Well, the big day is finally here. We will try to do a live blog tonight with live commentary as things unfold.

I want to take this time to urge all of you to put no stock in exit polls. I think they were proven to be a fraud in 2004, but nonetheless, some network(s) will use data from them at 5:30 to make broad inferences about the electorate.

Here's why they are flawed, based on some real stories from 2004.

First, the idea is fine, in theory. Go to some critical polling places, set up shop, and grab every fifth voter and ask them to fill out a survey, right?

Here are the problems:

1) Most of the exit pollers were graduate students, mainly women. I'm going to generalize here , but everyone remembers when they were in college, right? If someone said you had to be at a place at 6 am, but there was no one expecting you (or watching you) at that time....

you got in around 8, right? Right. No one's looking....

So a group of voters from 6 am to 8 am were missed. And the demographic of that group who votes at that time? Mainly working men. Who lean mostly Republican. So no exit voting was done of those voters.

2) So you are supposed to pull out every fifth voter. Fine. Remember, you are 21-23 years old working in the suburbs in an area you don't know.

You count people since your last survey...1...2...3..4..and then two people exit:
- a 27 year old stay at home mom, or
- a 55 year old man who is in a suit.

What do you have the most in common with? Who would you pick to interview? In most cases documented in 04, the 27 year old mom was picked if given a choice.

Most of the interviewers were democrats, it was found, and the natural human tendecny is to seek out people that "look" to be similar to you and have comparable life experiences/age/beliefs. We can't fault the interviewer for being reluctant to interview Republicans - but it was not taken into account.

3) The early results had to be reported to the network by 4 pm. So you've interviewed from 8 -4. Who are you missing from 4 pm - on? See the answer to #1 above.

4) The exit polls were set up in the traditional swing neighborhoods - not, of course, in all precincts - and the exit pools completely missed the GOTV operation in the ex-urbs that the Republicans set up, most notably in OH, FL, and other states.

This is how the exit polls showed that:
- Kerry would win PA by TWENTY POINTS (remember that?)
- The D's would be extremely close in VA, NC (they lost both by 10+)

This affects how the networks call races... I distinctly remember the networks not calling VA or NC when they closed ( Bush won both over 10 pts, as I said) but called MI and NJ 10 minutes after the polls closed (Kerry won MI by 3, NJ by 7).

The networks aren't biased - the polling procedure was.

I don't know what fixes were put in place for this year... but keep this in mind when reading any early tea leaves today.

(unless, of course, the news is good for the Republicans...then it's 100% accurate!)