In Fair’s model, signs point to Obama victory

Yale economics professor Ray Fair believes he can see the future — at least when it comes to presidential elections. And for Democratic presidential nominee Sen. Barack Obama, the future is looking good.

Using an algorithm he developed that forecasts the two-party popular vote of presidential races based on the current state of the economy, Fair is predicting an Obama victory over Republican Sen. John McCain — by a margin that he says may only grow between now and November as the current financial crisis worsens. Although critics of the model question an approach that takes into account only economics and not other forces, Fair’s “Presidential Vote Equation” has an impressive track record: It has correctly predicted the winner of the popular vote of every race since 1916 with only two exceptions.

The basic theory of voter behavior underlying Fair’s formula is the idea that a strong economy will favor an incumbent party, while a weak economy will benefit the opposing party. His equation — which he began to develop 30 years ago, around the time he joined the Yale faculty — uses economic indicators such as the rate of inflation and the per capita GDP growth rate to determine the percentage of the popular vote the incumbent party will receive in the upcoming presidential election.

Fair has only incorrectly predicted the winner of the popular vote twice, once in 1992 and once, retroactively, in 1960. After these two trials, he refined his algorithm.

In November 2006, Fair used his predictions about the state of the economy in 2008 to extrapolate that the Democrats would take 50.1 percent of the two-party vote in 2008, making the race too close for him to call. Yet, in the face of the recent turmoil on Wall Street, he said the economy is faring worse than his original data indicated — a good sign for the Democrats.

“If the economy really tanks this quarter — today isn’t such a good day for the economy — and the growth rate is quite low, then what I’m using is too optimistic,” Fair said Monday of his predictions for the 2008 economy.

Fair will not make his final prediction about the McCain-Obama race until numbers from the third economic quarter of 2008 are released in a few weeks. When those numbers are made public, Fair said, he will input economic conditions into the algorithm.

Until then, Internet surfers can use his “Compute your Vote Predictions” calculator on his Web site by inputting their own economic predictions to generate an outcome to the presidential race.

According to the calculator, Republicans better start hoping the economy shapes up before Nov. 4. Fair’s most recent prediction, using economic data from July 31, placed the Republican vote share at 48.5 percent.

School of Management associate professor Keith Chen, who has worked alongside Fair on other projects, said the equation is a “great guide” to understanding the history of the U.S. economy and how it can impact the political arena. But he said he thinks that Fair’s prediction could be unreliable in a close election.

“It distracts away from a lot of different aspects of the election including who actually gets nominated,” Chen said. “It doesn’t look at polling data. It just looks at who is in power and how the economy is doing.”

“He doesn’t have any evidence concerning this variable,” political science professor John Roemer said. “His correlations pit white candidates against white candidates.”

Fair has heard similar criticism before.

“Some people don’t like [the algorithm] because they think it ignores key things that shouldn’t be ignored,” Fair said. “The way we respond to that is we realize there are many things that affect voting behavior aside from what’s in the equation.”

Fair said these external factors are lumped into the “error term,” a measure of the likelihood that the test will produce an improper conclusion because certain external variables not factored into the final calculation affected the real-world result. For Fair’s model, that number is 2.5 percent, the difference between the actual value and the predicted value of the results.

“The interesting thing about this is that the error is as small as it is,” he said.

In addition to political races, Fair has applied econometrics to predict the effect of age on marathon times and the likelihood that an individual will have an extramarital affair. Fair said his ultimate goal is to get undergraduates excited about the potential applications of a field like econometrics.

Celia Stockwell ’11, an economics and math major who took “Introduction to Macroeconomics” with Fair last semester, said she thinks he is succeeding.

“Fair always found ways of making economic principles apply to interesting aspects of life,” she said.

Even some non-majors found what Fair had to say applicable. Elizabeth Deutsch ’11, who also took Fair’s macroeconomics course, said she was not initially interested in economics, but the class sparked her interest by showing her how economics can be used beyond the world of finance.

But some, including Fair himself, caution others not to take the predictions too seriously: The formula does not take the electoral college into account. Although Fair correctly predicted a narrow victory for the Democrats in the popular vote in 2000, the formula was not built to account for the effects of the electoral math that put Republican George W. Bush ’68 in the White House.

Although the equation predicted the wrong winner, the error of 1.3 percent — the margin by which Fair’s estimate diverged from the popular vote — was actually less than the model’s average error of 2.5 percent.

Comments

  • Sammy

    A model that factors in only the economy is likely to be unreliable for this election. There are so many factors that Americans consider and the polls are showing that McCain and Obama are tied. One can foreseeably create various models based on various factors giving various results. Human activity is hard to predict, especially when it is based on emotion and numerous other factors. I predict that McCain will win in 2008.

  • Not Sammy, Certainly Not Fair

    The problem is not that it predicts on only one variable. Conceivably, such a single-causation could exist. What makes we utterly incapable of taking this model seriously is that it in large part post-dictive, and has been modified (as stated) to "account" for where it went wrong.

    The fact that it missed 1992 further disturbs me, since there seem to be other negative economic indicators that should have helped Clinton — I wonder which Fair uses.

    Given that the article doesn not give details, it is hard to guess, but given a dozen elections, and say, half a dozen variables, that isn't really that convincing even a correlation, let alone causation.

    That said,

    Obama/Biden ’08 ;)

  • JE 10

    The problem, as I see it, is that this model suggests that it doesn't actually matter who the candidates themselves are. The Democrats could have run Chris Dodd and the Republicans could have run Duncan Hunter and the results would have been the same.

    That doesn't make any sense.

  • Ray Rules

    The model provides econometric evidence to support the clain that "It's the economy, stupid." Like any other scientific model, it's incomplete. For the exact specification and interpretation, please consult:

    http://fairmodel.econ.yale.edu/vote2008/index2.htm