Formula Forecasts Presidential Elections Months in Advance

By Daniel Gorelick

An accurate method to predict the winner of a presidential election has long been a dream of pollsters and political pundits. Now a Russian scientist and an American historian have a formula that, in the last six elections, has never failed.

The formula incorporates 13 variables, or keys, that can be classified as either true or false. If five or fewer keys are false, the formula predicts the incumbent party will win. The keys address economic growth, foreign and domestic affairs, social unrest and political scandal but ignore polling, advertising, news strategies and debates, techniques frequently used in American political campaigns.

Campaigns assume that advertising and political spin sway voters. The 13 keys suggest that voters are influenced more by the performance of the incumbent administration than by the candidates running for president.

If the outcomes of presidential races are as random as a coin toss, then the chance of predicting six consecutive elections is 1 in 64, or 1.6 percent.

CHANCE MEETING SPARKS COLLABORATION

In 1981, Russian geophysicist Volodia Keilis-Borok of the U.S.S.R. Academy of Sciences and American historian Allan J. Lichtman of American University were visiting professors at the California Institute of Technology. They discovered a common interest in American politics, especially elections.

Keilis-Borok realized the same mathematical techniques he was using to predict earthquakes, called pattern recognition, could be applied to presidential elections. Lichtman provided a historical insight: The incumbent party candidate, whether the sitting president or not, receives credit or blame for the record of the previous four years.

Combining pattern recognition with historical analysis, Keilis-Borok and Lichtman identified a series of statements whose responses, true or false, could be used to predict whether the incumbent party would win re-election. Thirty statements were tested to see whether their responses correlated with every election result since 1860. Some, such as those dealing with the stature of the vice president and whether the election occurs during a time of war, were omitted because they had weak associations with the outcome of elections.

Using the remaining statements, Keilis-Borok and Lichtman applied a pattern-recognition algorithm called Hammings distance and identified 12 that retrospectively predicted the popular vote winners of every election from 1860 to 1980 except the 1912 election.

With the help of statistician Harry N. Davey at the U.S. Bureau of Labor Statistics, they altered and expanded the variables to obtain a set of 13 questions that accurately predicted all the elections in this period.

Their analysis originally was reported in Proceedings of the National Academy of Sciences of the United States of America and expanded by Lichtman and Ken DeCell in the book The 13 Keys to the Presidency.

Beginning in 1981, Lichtman predicted the winner of every subsequent presidential election at least 10 months in advance. In November 1982 the United States was experiencing a severe economic recession and the incumbent Republican Party lost 27 seats in the House of Representatives, yet Lichtman - contrary to popular political wisdom at the time - correctly predicted Republican President Ronald Reagan would win re-election in 1984.

OTHER PREDICTION MODELS

Thus far Keilis-Borok and Lichtman's "keys model" appears to be more accurate than several other prominent prediction formulas.

Ray Fair, professor of economics at Yale University, developed a model based on inflation and growth in gross domestic product. Like the keys, Fair's model ignores polling data, but it failed to predict the results of the 1992 and 1996 elections. Economist Douglas Hibbs' model, based on changes in disposable income, incorrectly predicted the 1996 and 2000 elections.

One caveat of the keys is that they predict only the winner of the popular vote, which does not always agree with the electoral vote. Prior to the 2000 election, the keys predicted correctly that Democrat Al Gore would win the popular vote; however, Republican George W. Bush won the Electoral College vote and became president.

Unlike other models, the keys do not predict margins of victory or provide any estimate as to the percentage of voters who will support one candidate over another. Also, the keys are specific to presidential elections.

Lichtman and Keilis-Borok attempted to show different variables are important when analyzing House and Senate races. In 1989, they published a model to predict Senate elections, but this model has not been 100 percent accurate.

IMPORTANT IMPLICATIONS

If the keys truly reflect American voting behaviors, then polls, political consultants and strategists could be irrelevant. Presidential candidates, especially those predicted to lose, probably should concentrate on substantive issues and not play it safe, Lichtman said.

He added that the success of his formula also suggests presidential nominees should pick the best-qualified candidate for vice president, and not attempt to balance the ticket by choosing a vice president who appeals to a particular region or demographic group.

What do the keys say about the upcoming election in November? On October 4, 2007, Lichtman wrote that regardless of who heads the Democratic ticket, the Democrats will win the White House in 2008. "Only an unprecedented cataclysmic change in American politics during the next year could salvage Republican hopes."

On November 4 millions of Americans will head to the polls and the world will discover if the 13 keys can claim another election victory.

See "The 13 Keys to the White House ( http://www.america.gov/st/elections08-english/2008/October/20081002122328adkcilerog0.8765833.html&distid=ucs )."

For more on the U.S. presidential election, see the eJournal USA issue "Electoral College ( http://www.america.gov/publications/ejournalusa/0908.html )."

(