In the May 26 Washington Post, Robert Kaiser trotted out several of the finest minds in the academic business of election forecasting to announce Al Gore’s certain victory this fall over George W. Bush. “It’s not even going to be close,” the University of Iowa’s Michael Lewis-Beck was quoted as saying. But as Ira Carnahan pointed out at the time in these pages — joined by skeptical pieces in Slate and the National Journal — the election forecasters have a questionable track record. Their computer models can “predict” past elections but not future ones. Lewis-Beck, for instance, correctly foresaw Clinton’s 1996 win, but in 1992 he predicted a Bush victory.
Nonetheless, an unabashed Kaiser was at it again last week, treating Post readers to pro-Gore soothsaying from academic clairvoyants in town for the American Political Science Association’s annual meeting. Among Kaiser’s savants: Thomas M. Holbrook of the University of Wisconsin-Milwaukee. Like the others, Holbrooke’s model made its bones by correctly “postdicting” previous elections. Unfortunately, his model has correctly called only three of the five close elections since 1952. James Campbell of the University of Buffalo also pops up in Kaiser’s retread. His model, one 1993 analysis determined, contained an 18-point margin of error — big enough to drive a third party through. Also touted is the work of Christopher Wlezien, who in 1996 was only two-tenths of a point too low in his prediction that Clinton would win with 54.5 percent of the two-party vote. But, as the Post itself noted in 1996, Wlezien fine-tuned his numbers several times in the months leading up to the election.
THE SCRAPBOOK, in the same spirit, will issue its scientific predictions after hearing the early exit polls on Nov. 7. Meanwhile, the Bush campaign may have worries, but the political scientists’ forecasts shouldn’t be among them.
