Historically, the average error for Fair's model is 2.4 percentage points. Its worst performance was in 1992, when it was off by 5.1 points and mistakenly predicted re-election for the first President Bush, perhaps because voters' perception of the economy was worse than his model's measurement of actual economic conditions. Fair worries a similar perception gap or a disproportionate emphasis on foreign policy could come into play this year.
At the end of October, Fair will issue a revised prediction incorporating third-quarter economic numbers, but expects similar results.
Stocks tell a different story, predicting Kerry will win, barring a rally before Election Day, according to Jeffrey A. Hirsch, editor-in-chief of the "Stock Trader's Almanac." He notes the Dow Jones Industrials average is down from its 10,260.20 level on Sept. 3, the day after the last political convention, and its 10,080.27 level at the start of October. Generally, losses after the political conventions and during October have been bad signs for incumbents since around 1900.
"I think it's … economic market forces being affected by the policies of the current administration," Hirsch says. "If the market doesn't perform well, people vote with their wallets, or their statements."