I’ve estimated thousands of equations ever since I took my first course in econometrics back in 1966. Perhaps the most accurate of these is an equation developed by my students and me that forecasts the outcome of presidential elections.
This equation’s accuracy is reflected by the fact that it has correctly called the popular vote victory or loss for the incumbent party candidate in 15 of the last 16 elections. The first of those goes all the way back to 1956 when Dwight Eisenhower drubbed Adlai Stevenson for the second time in a row.
As for the one miscall in those sixteen elections, I’ll get to that later.
What makes the accuracy of our presidential forecast equation particularly impressive is that it does not rely on opinion polls or presidential approval rankings. Only two economic variables are included in the equation: the average percentage change in prices and the average percentage change in jobs during the year of the presidential election.
In our research, we not only tested the significance of different variables but also the timing of their influence. For example, it’s not, as one might expect, the change in prices or jobs during the fourth quarter of the election year that’s most telling. Multiple tests show quite conclusively that it’s the average change during the entire election year as compared to the prior year that best determines how people vote.
As shown by Chart 1, stronger average job growth during the election year clearly favors incumbent party candidates, while weaker growth leads to a “throw the bums out” mentality. For example, when job growth hit an impressive 4.7% in 1984, the incumbent Ronald Reagan trounced Walter Mondale by a wide popular vote margin of 18.3%. But when job growth declined by a half percent in 2008, voters blamed the incumbent party (Republican) candidate, John McCain, and cast their lot with Barack Obama by a margin of 7.4%.
While higher job growth helps an incumbent party candidate, high rates of inflation have the opposite effect. Many might argue that the incumbent Democratic candidate Jimmy Carter in 1980 lost in a landslide to Ronald Reagan because of the Iranian hostage crisis. While I imagine that didn’t help Carter’s cause, our equation suggests that his real problem with voters was that the average rate of inflation was raging at an astonishing 13.5% during the 1980 election year.
In addition to job growth and inflation rates as explanatory variables, we also found that incumbent presidential candidate benefits when seeking a second successive term of office. Evidently, the power of the presidency helps garner votes. That, however, doesn’t hold true for an incumbent political party that’s already occupied the White House for two successive terms and is seeking a third successive term.
Of the eight elections where an incumbent party vied for a three-peat, only George H.W. Bush won when he defeated Michael Dukakis following Reagan’s two tours of duty. All the rest lost.
Using all of these variables in our equation produces uncannily accurate forecasts of an incumbent party candidate’s popular vote margin of victory or loss in the popular vote. This can be seen in Chart 2, where the actual and forecasted lines are so close together, they can hardly be distinguished from each other.
Note that the presidential forecast equation correctly predicted the popular vote victories of Al Gore in 2000 and Hillary Clinton in 2016. This points out that the equation forecasts the popular vote margin—not electoral college votes.
Our recent Chapman Update Forecast suggests that the COVID-19 recession will not be kind to President Donald Trump’s reelection hopes. The forecast calls for the number of jobs in the U.S. to decrease 6.2% on average this year as compared to last year. Such a precipitous drop, the steepest in recorded economic history, spells doom for Trump. Plugging in our forecasted numbers in the equation points to a 10% victory for presidential hopeful Joe Biden in the popular vote.
Those who find this forecast hard to take may find solace in believing that voters won’t blame Trump for a COVID-19 induced recession. One might pin their hopes of another “Trump Miracle” on the belief that it’s different this time around.
Intuitively, the direction of jobs makes sense in affecting elections. Last year, the total number of nonfarm payroll jobs was 150.9 million. The Chapman forecast issued June 23 predicts 141.5 million jobs in 2020. That means 9.4 million potential voters have lost jobs this year. They are not in a forgiving mood to those in power, no matter what the reason.
Voters always blamed past incumbent presidents through those recessions may have had little to do with their leadership. Case in point: George H.W. Bush’s loss to Bill Clinton during the mild recession of ’92, a recession that actually ended before that election took place. Why would it be different this year? And if the COVID-19 recession is not perceived by voters as a Trump downturn, why then did pre-COVID polls in January show that 63% approved Trump’s handling of the economy versus a sharply lower 47% approval post-COVID?
Astute observers of Chart 2 might have noticed that the forecasted margin of victory for the incumbent party has deviated from the actual margin by larger margins of error since 2000. This may be the result of voters becoming more polarized in their political views and, as a result, voting along party lines rather than how they perceive the economy’s health. Nonetheless, the equation, while not quite as accurate since 2000, still correctly called the popular vote winner in every election.
One other factor to consider is that our equation points to a Trump loss in the popular vote by a wide margin of 10%. That margin might be whittled down for various reasons as the campaign heats up, but for Trump to overcome a big 10% deficit will require a whole lot of whittling.
Before I end, you might want to know the equations only miscall in the last 16 elections. That was in 1960 when the equation predicted a narrow victory of 2.9% in the popular vote for Richard Nixon over John F. Kennedy. Rather than a narrow win, Nixon actually lost by a scant 0.2%. Many historians, however, argue that Nixon would have won the popular vote as our equation forecasted if it weren’t for voter fraud in Illinois because of Mayor Richard Daley’s pull and in Texas because of Lyndon B. Johnson’s. So, I protest that result.
But that’s a whole other story.