The Phony Science of Predicting Elections

The Phony Science of Predicting Elections

The Phony Science of Predicting Elections

May 31 2000 3:00 AM

The Phony Science of Predicting Elections

Who'll win in November? The experts' guess is as good as yours. 

(Continued from Page 1)

9. Add loopholes. Political scientists claim that econometric models can explain elections because voting follows the same scientific laws year after year. Yet a prior Holbrook model adds a special variable for the elections of 1964 and 1972 to account for the "extremist" ideologies of Barry Goldwater and George McGovern. Fair (who, unlike most, tries to explain elections all the way back to 1916) adds a special variable for the three elections he believes were strongly influenced by war: 1920, 1944, and 1948. But when an election-year war doesn't fit the equation, as in 1968, Fair leaves that variable out.


10. Blame the lack of data. Holbrook told the Post that the 13 elections he analyzes are too small a sample, saying that with 30 cases he'd be much more confident in his model. This assumes that subsequent elections would clarify rather than complicate the range of data to be explained and the array of factors that might explain them. Analyzing elections 120 years apart using the same model is like trying to figure out whether Babe Ruth was a better hitter than Mark McGwire. They didn't face the same pitchers, they played in different stadiums, and the balls are manufactured differently today than in Ruth's day. Similarly, the transition from an industrial to a service economy, the change from one-earner to two-earner households, and the rise of the investor class make it a stretch to compare attitudes about the economy across generations.

At bottom, the models rest on three flaws. First, they assume what they're supposed to prove. They exclude factors such as the strengths of each candidate and each campaign, simply because political scientists don't know how to measure them. Campbell, for instance, decides not to incorporate the candidates' positions on issues in his model, since this factor is too "subjective" and "extremely cumbersome to calculate." In their efforts to provide "explanations" and "an understanding of what actually causes the vote on Election Day," the forecasters delude themselves: They can't predict or explain elections, because their models don't comprehend any aspect of human behavior that can't be quantified.

Second, the models boil down to truisms. They reduce elections to two independent variables: One measure of the health of the economy, and one measure of incumbent or candidate popularity. The values and coefficients they attach to these variables don't hold steady over time, but the principles do: People are inclined to vote with their pocketbooks, and popular candidates tend to get elected. Imagine that.

Third, the models separate objective conditions from the subjective advocates who present them to the electorate. As Russert put it to James Carville and Mary Matalin, econometric forecasts imply that what's going on in the campaign now is "meaningless," because, "It's the economy, stupid." But that phrase, coined by Carville in 1992, made the opposite point. He wasn't forecasting the election's outcome. He was reminding the campaign staff that the economy was a winning message for the campaign. The economy matters in part because candidates and campaigns make it matter, a subtlety lost on the number-crunchers.

Karl Eisenhower, a political science grad-school dropout, works for a software company.

Pete Nelson, a political journalism dropout, is a policy analyst at Resources for the Future. The opinions expressed here should not be attributed to Resources for the Future.