Electoral Vote Predictor 2004:   Kerry 252   Bush 279

Data in Excel format
Battleground states
Polling data     Key
Previous report
Next report


electoral college strong kerry Strong Kerry (95)
electoral college weak kerry Weak Kerry (105)
electoral college barely kerry Barely Kerry (52)
electoral college tied Exactly tied (0)
electoral college barely bush Barely Bush (30)
electoral college weak bush Weak Bush (66)
electoral college strong bush Strong Bush (183)
Needed to win: 270
Nov. 4 Preliminary election results RSS

News from the Votemaster

Well, it's all over. The people have spoken. Winston Churchill once said: "In a Democracy, people get the kind of government they deserve." So be it.

I am working now to clean up some odds and ends. The main new thing today is an improved pollster comparison. If you click on Compare the pollsters, you come to an expanded page comparing more pollsters in more states. The individual pollster pages have also been expanded, showing not only a map of the states the pollster surveyed, but also graphs of some of the data.

I started to fill in the standard spreadsheet used for all the polls with the final results accurate to two decimal places but then saw that not all precincts had reported yet, so I will delay this for a bit until more results are in. I am also working on more comparisons between the polls the election results.

I have a LOT of log data. For example, there is the page on hits from universities. If you are a professor of political science whose research specialty is polls and elections and have a track record to prove it, I might be willing to work with you. Wait a week before contacting me because I have to gather all the data from the 8 or 9 servers and there is a lot. I won't send anyone the entire logs because of (1) their size--about 6-8 GB per day, and (2) my desire to protect the visitors' privacy. However, I might be able to provide statistical information about how many hits from which foreign country and that kind of statistical data. Please bring your own programmer; I have done enough programming on this project already, believe me. In practice, what is probably going to work best is you give me a UNIX shell script that expects an Apache log file on standard input and produces some analysis on standard output. I then inspect the output to see if it contains any privacy sensitive data, and if not, I send you the output.

I am also looking for a couple of university political science departments interesting in hosting the data on their website for posterity although I will keep my HostRocket account open until after the 2008 election. Still, there is a lot of valuable data here and in the spirit of science I want to make sure it stays available for future researchers.

Two researchers in my Dept., Maarten van Steen and Guillaume Pierre, are working on analyzing the Internet traffic aspect of the data (as opposed to the political science aspect). They will publish the results in scientific journals and conferences.

Various people sent me mail saying that it is awfully fishy that the exit polls and final results were substantially different in some places. I hope someone will follow this up and actually do a careful analysis. Does anyone know of a Website containing all the exit poll data? If we go to computerized voting without a paper trail and the machines can be set up to cheat, that is the end of our democracy. Switching 5 votes per machine is probably all it would take to throw an election and nobody would ever see it unless someone compares the computer totals and exit polls. I am still very concerned about the remark of Walden O'Dell a Republican fund raiser and CEO of Diebold, which makes voting machines saying he would deliver Ohio for President Bush. Someone (not me) should look into this carefully. The major newspapers actually recounted all the votes in Florida last time. Maybe this year's project should be looking at the exit polls. If there are descrepancies between the exit polls and the final results in touch-screen counties but not in paper-ballot counties, that would be a signal. At the very least it could be a good masters thesis for a political science student. The Open voting consortium is a group addressing the subject of verifiable voting.

Having dealt with political science, now computer science. If you are a senior majoring in computer science and are seriously thinking of leaving the country due to the election results, you might be interested in my international English-language masters program in parallel and distributed computer systems. If you are a faculty member in computer science, I would be very grateful if you would go to that Website and download and print the poster (a PDF file) and pin it to a bulletin board where potential students might see it and mention it in any classes you teach to CS seniors. Thank you.

Projected Senate: 44 Democrats, 53 Republicans, 1 independent, 2 tossups
This site has far more about the election than just the map. See the Welcome page for more details.

-- The votemaster

WWW www.electoral-vote.com

Statistics Collector (via University of Kentucky)