Ray Fair writes:
Presidential Vote Equation--October 29, 2004: The predictions of GROWTH, INFLATION, and GOODNEWS for the previous forecast from the US model (July 31, 2004) were 2.7 percent, 2.1 percent, and 2, respectively. With the release of the NIPA data on October 29, 2004, all the actual economic data are available for the vote prediction. The actual values (as of October 29, 2004) of GROWTH, INFLATION, and GOODNEWS are 2.9 percent, 2.0 percent, and 2, respectively.
Given that the actual economic values are close to the values used for the previous vote prediction, the current vote prediction is little changed. The new economic values give a prediction of 57.70 percent of the two-party vote for President Bush rather than 57.48 percent before.
I suspect Fair's vote equation is contaminated by the inevitable tendency to indulge in specification searches, but if you take his standard error of 2.37% seriously, then Bush--who got not 57.7% but 51.5% of the two-party vote--undershot his expected vote share by 2.6 standard deviations, a Bush performance so bad (or a Kerry performance so good) that you would expect to see it only 1 in 200 times.
Posted by DeLong at November 3, 2004 09:28 AM | TrackBack