Predicting elections every four years has become a pastime — one that has seen much criticism and debate. Forecasting elections based on economic fluctuations became popular in the 1970s, and was hotly debated matter. Yale economist Ray Fair, who is renowned for his macroeconomic voting model, was on that bandwagon and has since authored several books on the subject.
In his book published just this year, “Predicting Presidential Elections and Other Things,” Fair goes in depth about his “tools of social sciences and statistics” to predict the popular vote prior to presidential elections. Like many forecasters, Fair uses information from past elections to predict the outcome of an election that has not yet taken place. “Election results can be used in a systematic way to decide which stories have merit,” he wrote.
He also suggests that the determinants of voting behavior are not limited to economic fundamentals, but could involve non-economic factors as well. Relying on gross domestic product, inflation, and incumbency metrics, Fair’s theory reveals a positive relationshipbetween growth rate in year of election and incumbent vote share.
To determine whether the incumbent does well in a good economy and bad in a bad economy, Fair proposes to test this theory by testing the GDP and the two-party vote share the candidate receives. For example, the economy experienced 3.3 percent growth in 1996, and incumbent Bill Clinton received 54.7 percent of vote share, defeating Bob Dole. However, in 2008, growth was negative 2.3 percent, and the incumbent Republican party received only 46.3 of two-party vote share, losing to Barack Obama.
However, I have found a few bones to pick with Fair’s literature.
1. Interestingly enough, Fair is quite frank in explaining that the aim of his model is not to explain who wins the election. Rather, the theory is judged by how close the predicted values of the two-party vote share are to the actual values, by the size of the errors — and there are always errors.
Quick reminder: His book is called “Predicting presidential elections … ”
Also like many presidential election forecasters, Fair can only hope for small errors in his predictions, but accuracy is not achievable.
2. Later, in addressing the role of the electoral college in his theory (hint: there is no role), he again says “The aim of the analysis is not to explain how many states go for one party over another.” As I pointed out in my previous post, what is the significance of a voting model if it reflects only the popular vote?
3. In his book, Fair says “Not counting 2000, three elections were predicted incorrectly as to the winner: 1960, 1968, and 1992. … even though the winner was predicted in these two elections, the errors are small, so in this sense, the elections were predicted well.”
I’m can’t imagine the losing candidates would agree with this inference.
4. He does raise the question of whether a campaign matters, and he says that they are “likely” to matter, but does not say in what way. He also says the the nomination of the person is also likely to matter, but none of these factors made it into his model. It is apparent that he has not researched these possibilities in as much depth as, say, Lynn Vavreck.
In my studies this semester I have been exploring Vavreck’s theory. Her model is a multidimensional approach that not only explains why candidates win elections, but also how voters make their decisions. It has been very interesting, and at times confusing, to see how economists and scientists using the same data can draw completely different conclusions. It seems as though Vavreck’s wide and inclusive use of that data contrasts with Fair’s narrow and selective system.
While election econometrics is interesting, I am not convinced that numbers, statistics and bell curves can explain why voters consciously vote the way they do. In the 1970s, political scientists and economists were focused on drawing connections between economic fundamentals and incumbent presidents. Forty years later, many voting theories still put heavy emphasis on these two factors. In some cases, such as Vavreck’s, long strides have led to a more complete picture of voter behavior, while others are still beholden to the initial findings that only leave me with more questions than answers.