"Almost all the [election] models incorporate some measure of economic conditions . . . as well as a measure of the current president's job performance," both factors that favored Gore in the long-range predictions, explains Alan Abramowitz of Emory University. "We may have missed out on the huge dichotomy in the public's opinion of President Clinton, in terms of job performance verses personal evaluation." Abramowitz, who was one of the seven forecasters at the APSA meeting, cites third party candidates as another factor complicating political models.

Regardless of the outcome, Abramowitz says that he and his political science peers view their work as an academic pursuit, and won't be terribly disappointed if they miss the mark by a few percentage points, even if they fail to predict the ultimate outcome. "I don't think it would undermine the validity of the enterprise [if the predictions were wrong] because we know there are factors that influence the election that are outside the models," such as President Clinton's unique impact. Instead, according to Abramowitz, a failure of their predictions could give researchers valuable fodder for developing and refining future models.

If the failure of political models in this year's election, swings in polls over the past several months, and even the candidates perpetually redefined opinions seem a bit chaotic, there could be a very good reason for it. David Meyer, a mathematician at the University of California at San Diego who studies voting and other real world behaviors using theoretical tools derived from physics, sees parallels between electoral results and the inherent unpredictability of chaotically evolving systems. "If there's a small change in the outcome of one election," says Meyer, "then several elections down the road, things maybe very different than they would have been had the [initial] change not occurred." While Meyer's hypotheses clearly suggests that predicting the results of elections in distant years is a futile exercise, it's possible that even the act of polling, which is in effect a mini election, could cause unpredictable changes in the few months leading up to an election.

Meyer also considers the campaign process from the other side, that is, he ponders the effect polls have on candidates' opinions. "The series of decisions a candidate makes in response to polling information can explain why their positions are sometimes hard to pin down," Meyer explains. "We do an analysis," says Meyer, "where we find the vagueness of [a] candidate's position to be described by a probability distribution over the issue space, if you like." Which, in translation, means that Meyer and his coauthor Thad Brown have found correlations between the behavior of shifty politicians and the mathematics of physics.

While Meyer's approach is not designed to predict election outcomes, it may eventually help us to understand why politicians are so hard to understand, and perhaps bring some measure of comfort to the beleaguered political scientists whose models seem to be at odds with the voting public.

Note: If no author is given, the source is cited instead.

• Computer Modeling
• Mathematical Modeling
• Statistics

• Experimental economics
• Mathematical model
• Global climate model
• Microeconomics
• Game theory
• Macroeconomics