The Presidential Vote Equation: 1996 Update
The update of the presidential vote equation through the 1996 election is described in Fair (1998). The date of this update is November 6, 1998. This equation was used to predict the 2000 election. The specification of the equation for this update was not changed from the previous one (updated through the 1992 election and used to predict the 1996 election). The equation was simply reestimated using updated data and the one extra election observation.

The first prediction of the 2000 election using the update was made on November 6, 1998. This is described in November 6, 1998, prediction. The predictions that followed are then:

January 30, 1999, prediction
May 1, 1999, prediction
November 5, 1999, prediction
January 29, 2000, prediction
April 28, 2000, prediction
July 31, 2000, prediction
October 27, 2000, prediction

During this period users were allowed to use the equation to make their own predictions. See Compute your own vote prediction for how this was done.

The actual outcome was that Gore received 50.3 percent of the two party vote. The last prediction on this site (October 27, 2000), which used actual values for all the economic variables, was that Gore would receive 50.8 percent of the two party vote. The error is thus 0.5 percentage points. The standard error of the equation is 2.15 percentage points, and so the actual error is well within one standard error. The presidential vote equation thus did extremely well.

The previous two vote predictions on this site (July 31, 2000, and April 28, 2000) also had the Democrats with 50.8 percent of the two party vote---you can click these two predictions above. It turned out that the predicted values of the economic variables that were used for these two vote predictions (which were predicted values of the economic variables from the US model at the time) were very close to the actual values that became available in late October. Therefore, the July 31, 2000, and April 28, 2000, vote predictions were essentially the same as the October 27, 2000, vote prediction. So as early as April 28, 2000, this site was predicting what actually happened with an error of only 0.5 percentage points. And this was before we even knew who the two candidates were!

In January 2001 I wrote a brief article discussing what the vote equation (not yet updated through the 2000 election) had to say about President Bush's chances in 2004. See January 15, 2001: Bush's Chances in 2004.