Post-mortem--November 3, 2004

Real-Time Predictions of the 2004 Election

November 2001 1.5 3.0 3 56.9 51.6 5.3
November 1, 2002 2.0 1.9 1 56.3 51.6 4.7
January 30, 2003 2.0 1.9 1 56.3 51.6 4.7
April 25, 2003 2.0 1.9 1 56.3 51.6 4.7
July 31, 2003 2.4 1.8 1 56.7 51.6 5.1
October 31, 2003 2.4 1.9 3 58.3 51.6 6.7
February 5, 2004 3.0 1.9 3 58.7 51.6 7.1
April 29, 2004 3.2 2.0 3 58.7 51.6 7.1
July 31, 2004 2.7 2.1 2 57.5 51.6 5.9
October 29, 2004 2.9 2.0 2 57.7 51.6 6.1

The first prediction is from Box 4-2, page 65, in Predicting Presidential Elections and Other Things. This prediction uses the 1996 update of the vote equation (the one discussed in the book), not the 2000 update. The second prediction is from the paper on this site: "The Effect of Economic Events on Votes for President: 2000 Update." This prediction uses the 2000 update, and all the remaining predictions use the 2000 update. The economic values for the last prediction are actual values as of October 29, 2004.

One noticeable feature of the vote prediction is that it did not change much over the three year period. As far back as November 2001 it was predicting that President Bush would be hard to beat if the economy were moderately good or better. Compare this stability to the large fluctuations in the polls even in the last few months before the election!

The other noticeable feature of the vote prediction is that it substantially overestimated Bush's actual vote share. The final error was 6.1 percentage points. (This error is likely to change at least slightly as the current economic data are revised.) The vote equation says that with Bush's incumbency advantages and with the moderately good economic variables, he should have done much better than he did. There are many reasons one can think of as to why Bush did not do better. Many of these are discussed in the "Note to the Media: July 31, 2004," which is on this site. The problem regarding testing is that there are many reasons and one observation. There is no easy way to test which reason might be right. My personal view is that were it not for Iraq, Bush would have come close to the equation's prediction, but again this cannot be tested.

It is possible to use the current vote equation to make a prediction for 2008. There will be no incumbent running again (PERSON = 0), and the Republicans will have a negative duration effect (DURATION = 1). If, say, GROWTH is 3.0, INFLATION is 3.0, and GOODNEWS is 2, which is a moderately good economy, the vote prediction for the Republicans is 50.1 percent, a dead heat. So the main message for 2008 is that the election will be close if the economy is moderately good. It would take a quite strong economy for the equation to predict a comfortable Republican win, and it would take a quite weak economy for the equation to predict a comfortable Democratic win. The Democrats clearly have a much better shot in 2008 than they had in 2004 according to the equation.