September 15, 2008
The Votemaster is BackA sign that I've been blogging for a while: I can now link to entries from the last US presidential election. For example, I noticed recently that the Votemaster, who in 2004 ran a site tracking poll results by state, is doing it again this year. In 2004 the Votemaster was revealed to be Andrew Tanenbaum, author of several books on operating systems; I assume it is him again.
Tracking results by state is what you want, of course; national polls don't really matter because the election isn't decided by a national vote. An equivalent exists (as it often does) in baseball. Imagine if you spent all year looking at difference in total runs scored vs. allowed, but then the actual playoff results were based on this notion of grouping the runs for and against into 9-inning games. Looking at run differential for the league, you might conclude that Toronto was tied for second place, not 9 games back in the wild card race. It's the same thing with national polls vs. the per-state results that actually matter on election day.
If you look at the results for today it's almost a tie, with McCain ahead 270 to 268. It looks like, as in 2004, the race will come down to who can win Ohio. From this graph of votes vs. time we can see that Obama had a large lead throughout the summer, but has lost it in the last week.
Posted by AdamBa at September 15, 2008 09:16 PM
One of the better State by State Polling trackers is 538 (http://www.fivethirtyeight.com/)
Posted by: KnowOne at September 16, 2008 05:12 AM
Commentators seem to agree that you have to subtract 7% from Obama's total to account for the people who will not openly admit that they won't vote for a "black" (he is as much white as black). This number has been verified, at least approximately, in a number of races. Even exit polls are off by that amount. I am afraid Obama is toast.
Among other things, many years ago there was a large push to increase the number of voters. Now, they are concentrating on how to disqualify the wrong kind of voters. A good friend of mine who lived in Oberlin, OH, in '04 had to wait four hours to vote. The county had come in a few days before the election and moved a number of voting booths to rural areas where they weren't needed. Of course, Oberlin is home to Oberlin College and all those students. Could it be they didn't want them voting? And there were reports that people had to wait up to 10 hours in black neighborhoods in Cleveland. You will never convince me the election wasn't stolen right there in Ohio.
Indiana has a new voter id requirement that for all practical purposes restricts voters to people with driver's licences. Yes there are alternate forms of id but you have to drive to get them in rural areas. I just read that 77% of adults without driver's licences were black (and 17% white). More to the point, a slight majority of blacks in Indiana did not have a licence. Of course, this is all perfectly acceptable to the Supreme Court.
Posted by: marble chair at September 16, 2008 02:52 PM
Is that 7% nationally or in every state? I assume it is unbalanced so that in some states he loses 14% and in some he loses none.
I mean, if you take 7% off of New York, then that puts McCain in front. Are you seriously suggesting that New York won't go Democrat?
Posted by: Adam Barr at September 16, 2008 09:44 PM
I think the Bradley effect (that Marble Chair refers to above) hasn't been shown to reliably exist in this election. In fact in some (more heavily black) states there is a "reverse Bradley effect", i.e. more support at the ballot box than in the polls.
In any case 7% sounds way way too high. If that were true InTrade (and others) would not be trading as if the election essentially tied at the moment.
Posted by: KnowOne at September 17, 2008 07:42 PM
It doesn't matter how people vote, what really matters is who counts the votes.
In case you have missed HBO documentary "Hacking Democracy" you should really watch it.
Posted by: Ivan at September 19, 2008 03:04 AM