Electoral Vote Predictor 2004: Kerry 270 Bush 248
News from the Votemaster
Mondays are always quiet and today is no exception. An Opinion Research poll in Arkansas breaks the tie there and puts Bush ahead by 9%. I don't think Kerry really has much of a chance in Arkansas. The south, except for Florida, looks pretty solid for Bush, although there might be surprises in one or two border states.
Zogby did a large (N = 1216) telephone poll Oct. 7-9 (thus, after the second presidential debate) and found the race to be a statistical tie, with Kerry at 46%, Bush at 45%, Nader at 0.9%, Cobb at 0.2%, Peroutka at 0.2%, and Badnarik at 0.1%. The rest are still undecided. I guess they are waiting for the third debate, on Wednesday. Some people like to collect all the data before coming to a conclusion. Interestingly, Zogby also found that among newly registered voters, Kerry holds a 5% lead. Given the millions of people who registered for the first time this year, new voters (along with the millions of overseas voters) could be a serious factor.
After many attempts to turn the predicted map into something believable, I have concluded there are too many outliers, too much noise in the data, too many differences between the pollsters and too many voters who can't make up their minds, I have given up and removed the predicted map. Look at the state graphs to see the nature of the problem. The data are not converging. There is no pattern here. I tried different time intervals and different weighting schemes, but nothing seems reasonable. The quality of the data just isn't good enough. I think the current map is a far better predictor than the 'predicted' map ever was. Chalk one up to experience.
However, replacing the predicted map on the menu below the map is a new page called Compare the pollsters. This page shows maps and charts per pollster so you can see how they differ. You be the judge who has their thumb on the scale. I will drop a polling firm only when a mainstream media outlet has caught them red-handed. Some people have asked for a method to make their own maps selecting only pollsters favorable to their candidate. While this is theoretically possible, the computing load it would place on the server would be too much. That could be reduced by my computing two maps in advance: one where Kerry wins and one where Bush wins, but I am not sure of the value of such maps. On the More data page there are links to two sites where you can concoct your own data and make your own maps.
electoral-vote.com |
