Reply #54: you cannot say this w/o more info [View All]

I see your point, but that would only be accurate at best with the east coast results that would be in the 7:30 EST window, and it would be highly unlikely that 30% of the vote was in 30 minutes. However, if you open that time window to the Central, Mountain, or Pacific Time you are far open to a greater percentage of the total vote being missed.

Here is how I see it, you have a large number of voters casting votes before they go to work, or after they get off. You may have a trend for Kerry throughout the day that gets swamped when all the after work voters are Bush supporters.

What I would do is look at the adjustment on 11/3 to the Eastern set, and see how that changes. What you may observe as a Kerry trend may in fact reflect the east coast only.

What this all means is that Mitofski seems to have an east coast bias to his sampling methodology, and that may be why he has to correct his data. In other words, he has an outdated (read crappy) sample design (built from the east coast was all days of the 1960s), unless his 7:30 exit poll data is within each time zone (that is a big assumption on my part that it is not!); because he would stand to miss significant voter movement in the more populous states of California, Texas, Michigan, Ohio etc.

I really wish the source had released this earlier, because I think we now had better re-think how exit polling as conducted in the United States may give an indication of the honesty of elections. Consider this the biggest of bum steers.

Mike