Predicting the Next President
With the 2016 presidential elections behind us (for better or worse), most media organizations and pundits predicted the outcome incorrectly. One economist, however, did make the correct call. In fact, Cowles Research Staﬀ Member and John M. Musser Professor of Economics, Ray Fair, has quite the track record for predicting the presidential outcome with his voting formula.
It is human nature to speculate what the future holds and many people like to make predictions. What’s more, there are about as many formulas arriving at a prediction as there are predictions. Some may be based on scientiﬁc and analytical data, while some may be purely speculation on other factors.
Take for example sporting events; people are always trying to pick which team will win the Super Bowl, World Series, or World Cup. The same holds true of the stock market: Will it be bearish or bullish? And elections are no exception when it comes to predicting the outcome; people instinctively want to know who will be their next leader.
The 2016 presidential election results proved to be surprising to many people, however, particularly pundits. According to the Pew Research Center, pollsters predicted Hilary Clinton’s chances of winning to be at a whopping 70-99%! As it turns out, the actual poll numbers according to research by Theodore de Macedo Soares, show a 2.4% margin of discrepancy between exit polls in favor of Donald Trump, not the expected 3.2% margin favoring Clinton.
While political scientists and pollsters were left scratching their heads and will continue to pour over data develop case studies, Professor Ray Fair is once again relishing in a correct prediction using his own model which he developed nearly 40 years ago.
Fair’s method looks at fundamental forces driving people’s voting behavior rather than asking people who they plan to vote for, or how and why they came to make their decisions. Simply stated, his approach, “tries to explain what motivates people to vote the way they do.” He elaborates in his paper “Econometrics and Presidential Elections” by writing, “The general theory behind the model is that a voter evaluates the past economic performances of the competing parties and votes for the party that provides the highest expected future utility.”
Rather than relying on typical exit poll questioning, Fair uses four conditions to determine voter preferences: if a president is seeking a second term; the duration a party has controlled the White House; a persistent bias (albeit slight) in favor of the Republican Party; the state of the economy, particularly the rate of growth of output (GDP) and the inflation rate of the previous 15 quarters leading up to the election.
Each condition can be broken down for further clariﬁcation. His model states that nominees running for a second term tend to have a better chance of winning. If one party has controlled the White House for a long duration (e.g., two terms), this is seen as less favorable for the incumbent party. Fair’s research also shows voters tend to lean in favor of the Republican Party which he equates into his model. And lastly, he looks at the state of the economy during the ﬁrst three quarters of the election year to see where output growth and inflation stand.
Had the Republicans nominated a more main stream candidate, they may have done much better—much closer to what the equation was predicting.
As it turns out, the ﬁrst three criteria were not in Hilary Clinton’s favor. As for the last condition, while inflation was low in 2016, the GDP was historically weak despite a slight up-tick in the last quarter. Thus, the conditions pointed to a Trump victory.
Although Fair predicted the Democrats would receive less votes, his results were not as close as anticipated. Since November 2014, his model had favored the Republicans, predicting the Democrats to receive 44 percent of the two-party vote. Clinton, however, ended up receiving a lot more votes, which was actually closer to 51.1 percent. That’s an error of 7.1 percentage points. More surprising, this was nearly twice the error average Fair saw over the last 9 elections which averaged 3.53 percent points.
Fair cannot say with certainty why he had such a swing in the error, but he attributes the discrepancy to Trump’s personality. “Had the Republicans nominated a more main stream candidate, they may have done much better—much closer to what the equation was predicting,” says Fair. “The election was theirs to lose because of the economy and the duration eﬀect, and they almost lost it!”
Fair says he became interested in voter behavior in the early 1970’s while teaching at Princeton University with Orley Ashenfelter who was looking for an econometrics problem assignment for his students. According to Fair, “the subject matter is interesting, the voting equation is easy to understand, and the econometrics oﬀers practical problems.” While Fair is not aware of his model used for other elections, e.g., gubernatorial races, he is not alone in developing presidential prediction formulas. In fact, American University History Professor Allan Lichtman, created the Keys System and has successfully predicted nine presidents since 1984.