Behind the high-profile humiliation of journalists and pollsters who wrongly predicted the outcome of this month's US presidential election is an entire faculty of embarrassed academics whose projections fell even shorter.
Using models based on previous elections, political scientists at US universities had almost universally pronounced Al Gore the likely winner over George W. Bush, some by a margin of as much as two to one. The actual result was nearly an even split.
Particularly embarrassed was a panel of top scholars who convened during a meeting of the American Political Science Association in Washington just after the two major parties' nominating conventions and said that Gore would easily prevail over a better-than-expected Bush.
In hindsight, the academics blame unforeseen variables such as the two candidates' very different personalities, disclosures about Bush's arrest for drunken driving and the effect of Green Party upstart Ralph Nader.
"Election projection is fraught with difficulties inherently," said Lee Sigelman, a professor at George Washington University who was on the panel. "People are trying to take a very small number of cases and build accurate forecasting models."
Those models are based on the 13 postwar presidential races, a comparatively small scientific sample, Dr Sigelman points out. "When you are operating in that kind of context, you become the creature of a few data points, and any idiosyncratic factor in a particular campaign can throw your models off substantially," he said.
Political scientists probably relied too heavily on the state of the economy to find for Gore, he said. Also, he said, "intense media scrutiny certainly throws a great deal of fluctuation into the mix", something that has been true only in the past five or six US presidential elections.
Despite this, many political scientists were bolstered in their beliefs when Bush surged ahead in September, closer to where they had projected he would finish. "It was like vindication of the theory right there," said Robert E. Erikson, a Columbia University professor who was also a panel member. "At the time, the world was consistent with the panel's expectation. We were not looking for things to go wrong."
But several variables quickly cropped up to throw off the forecasts, he said, most notably the Nader campaign. "Nobody in the summer took it into account that I know of," Dr Erikson said.
Another political scientist said the discipline relied on a methodology called the "rational choices" school of thought, which fails to take account of "irrational" decisions. "All these models suggest is that if you know the state of the economy and you know, say, the presidential approval rate, which is a crude indicator of the state of the world, then you can predict the vote with a certain degree of confidence, but also a certain degree of error," he said.
"This ain't an exact science," Dr Sigelman added. "You're holding your finger up in the wind."