Friday, September 03, 2004

Darn the luck! 

At a panel discussion for the American Political Science Association, six of seven papers presented predicted victory for President Bush. Scott Armstrong, Alfred Cuzan and Randall Jones operate a site which summarizes the seven models and their predictions. The one that does not predict victory necessarily for Bush, by Michael Lewis-Beck and Charles Tien, has him ahead but within a standard deviation of tied with Kerry, ergo, a statistical dead heat. Even the more pessimistic models, nevertheless, would put the probability of a Bush victory at 0.75 (i.e., a 3-to-1 favorite.) One of the models presented was Ray Fair's.

According to a report in the Chronicle of Higher Education (subscribers only), the audience of this panel discussion was a bit dismayed.
During the question-and-answer period, several speakers wondered whether the models effectively capture public sentiment about the Iraq war and other foreign-policy debates. Thomas E. Mann, a senior fellow at the Brookings Institution, asked: "Does anyone in this room doubt that if the president had chosen not to go into Iraq militarily, and instead had maintained a broad coalition at home and abroad, to finish Afghanistan and focus on Al Qaeda, that he would be sweeping to victory? ... If we don't take into account the real politics of this election, then I think we're missing something."

Most of the panelists replied that their models do in fact include public sentiment about Iraq. All of the models incorporate polling data about the candidates' (or at least President Bush's) public-approval ratings. Despite the economic variables built into the models, they said, it is a mistake to believe that they are narrowly driven by economics.

In response to another challenge from an audience member, the panelists asserted that their quantitative models have broad scholarly value and do not serve only to titillate political professionals and the news media. The models have revealed that presidential elections should be understood almost entirely as a referendum on the incumbent party, said Alan I. Abramowitz, a professor of political science at Emory University.

A few of the panelists on Thursday expressed dismay at their own predictions. Their
anguish brought a smile to the lips of Mr. Campbell, who said that he is one of the few Republicans in this subfield. He said that he had been personally unhappy when his 1996 and 2000 models found that Democrats would win the popular vote, and he is bemused to see his colleagues' discomfort in this round. "It's time that the Dems took one for science," he said.