Election Results Post-Mortem from Ray Fair
In a follow up to the post from Nov. 3, Ray Fair writes:
The vote equation did well in predicting the 2008 election. The following chart lists the nine predictions that were made from November 1, 2006, through October 30, 2008. The most important prediction from the point of view of judging the vote equation is the last one, which uses the actual economic values (actual as of October 30, 2008). The two-party vote was 53.0 percent for Obama and 47.0 percent for McCain.
Nine Predictions from the Vote Equation
Dem.--Rep. GROWTH INFLATION GOODNEWS ERROR
Actual 53.0--47.0 0.22 2.88 3
Nov. 1, 2006 53.5--46.5 1.8 3.6 1 -0.5
Jan. 31, 2007 53.4--46.6 1.7 3.4 1 -0.4
Apr. 27, 2007 53.2--46.8 1.9 3.3 1 -0.2
July 27, 2007 52.0--48.0 2.2 3.5 2 1.0
Oct. 31, 2007 51.9--48.1 2.2 3.3 2 1.1
Jan. 31, 2008 52.0--48.0 1.8 3.1 2 1.0
Apr. 30, 2008 52.2--47.8 1.5 3.0 2 0.8
July 31, 2008 51.5--48.5 1.0 3.0 3 1.5
Oct. 30, 2008 51.9--48.1 0.22 2.88 3 1.1
One feature of the predictions is that they did not change much across time. On July 27, 2007, the NIPA data revisions led to one more good news quarter, as did the July 31, 2008, data revisions. Other things being equal, each additional good news quarter leads to an increase in the incumbent (Republican) vote share of 1.075 percentage points. The predictions for the Republicans varied from 46.5 to 48.5 across this two-year period.
Another feature of the predictions is that they are all fairly accurate. The errors are well within the standard error of the equation of 2.5. In the final analysis the error was 1.1. In general, even two years before the election the equation was giving a fairly accurate view of the election outcome.
The website for the vote equation is: http://fairmodel.econ.yale.edu/vote2008/index2.htm
Comments