close
close

Association-anemone

Bite-sized brilliance in every update

Which election prediction models can you trust?
asane

Which election prediction models can you trust?

play

WASHINGTON – As voters look to polls and political analysts to find out who might win the presidency on Tuesdaya feud between two of the nation’s leading election forecasters, Allan Lichtman and Nate Silverwill soon be put to the test.

Lichtman, a professor at American University who has correctly predicted nine out of 10 of the last presidential elections, predicted a victory for Vice President Kamala Harris.

Silver, the statistician and pollster who founded FiveThirtyEight, recently wrote in the New York Times that the race is a virtual tie, but his “gut” tells him former President Donald Trump will probably prevail.

Lichtman and Silver confront methods

The pair has they argued on social media about the validity of the respective methods.

In September, Silver questioned whether Lichtman correctly evaluated the “13 keys” he uses to project election results, arguing that the professor’s system actually favors Trump. Lichtman countered that Silver, whose background is in economics, “was not a historian or a political scientist” and had been wrong in the past.

“At least 7 of the keys, maybe 8, clearly favor Trump. Sorry bro, but that’s what the keys say. Unless you admit they’re totally arbitrary?” Silver posted on social media.

So whose prediction is more accurate? And how do they come to these conclusions in the first place?

Forecasting approaches

Lichtman devised the metrics he uses for his election forecasting more than three decades ago with the help of a Moscow-based specialist and mathematician named Vladimir Keilis-Borok.

The system, called “13 Keys to the White House” uses—you guessed it— thirteen true or false statements rooted in historical analysis about the state of the country, parties and candidates to determine who will win.

It includes questions about whether there is a third-party challenger, whether the “White House party is avoiding a primary contest” and whether either candidate is charismatic.

The method doesn’t take into account how campaign messages or major events like debates influence voter sentiment. Lichtman often makes his assessment months before an election and does not change it unless major foreign policy events occur.

When six or more of the statements are true, the challenging side is expected to win. When five or fewer are false, the incumbent is expected to win. In 2024, Lichtman said at least eight of the keys favor Harris.

But Silver uses an entirely different strategy and set of data points to examine the state of the election.

He builds probabilistic statistical models based on national and state polls, economic data points, likely voter turnout and other factors. The model it also adjusts for discrepancies in the polls it collects and weighs more heavily on those it deems more trustworthy.

Prediction records

Lichtman has correctly predicted the outcome of nine of the last 10 presidential elections, dating back to 1984. The one he got wrong? The 2000 presidential race in which George W. Bush defeated Al Gore.

Silver gained national recognition in 2008 when his statistical model correctly predicted the outcome of presidential elections in 49 of the 50 states. His model has since predicted the outcome of the 2012 and 2020 presidential races. During the 2016 election, Silver’s model suggested a likely victory for Hillary Clinton, but gave Trump about a 30 percent chance of winning — much higher than most other preachers.

Which model is better?

That depends on who you ask.

Thomas Miller, director of Northwestern University’s data science program, argued that both Silver’s and Lichtman’s strategies are “wrong in different ways.” Miller created a election forecasting system own, combining 60 years of historical analysis and data from the Predict It betting market.

He suggested that Lichtman’s model does not account for how campaign messages and major events change public sentiment in the final months of an election.

“According to Lichtman, none of the campaigns really matter. The messaging doesn’t matter, the positioning doesn’t matter … because everything is predetermined, in a way, by history,” Miller said. He also questioned whether the economic metrics Lichtman uses, which look at the gross domestic product of USA, precision target perceptions about the economy.

This year, for example, inflation is a major issue for many voters. The US economy is performing relatively well, but voters don’t necessarily feel it.

But Lichtman rejected those claims and said his economic analysis is objective and rooted in history dating back to 1860. Each key is well-defined based on that analysis, Lichtman said. He claimed that the absence of campaign events in his keys is one of the reasons they have been so successful.

“What some say is the weakness of the keys … is the strength of the keys because they look at the fundamentals, not the ephemeral events of the campaign,” Lichtman said. He said the structural model reflects how US presidential elections actually work.

Miller also saw flaws in Silver’s approach, namely that it relied too heavily on polling data, which varies and is fallible. If the polls are inaccurate, Silver’s predictions will be inaccurate.

Weighting polls by which groups of people are more likely to vote can also be tricky, Lichtman said. Polls can and have underestimated the number of Democrats and Republicans showing up for election, for example.

David Wasserman, an election analyst for The Cook Political Reportsaid that, despite the variability, he found Silver’s approach “more methodologically rigorous”.

“Lichtman is comically overconfident and doesn’t acknowledge the subjectivities of his method,” Silver said in late September, “but you’ll learn a legit lot about presidential elections by reading his work, and at least he’s putting himself out there. making testable predictions’.

Wasserman said he believes Silver’s approach is better for “conveying to the public where the election is,” in part, because he “recognizes that there is uncertainty inherent in polls and future events.”

“I believe that campaigns matter … and that the candidate’s choices affect the way voters think,” he said. “I put more into Silver’s approach because it can take those factors into account.”

In essence, however, the models are completely different.

Where Lichtman’s model looks at established patterns of past elections to predict future presidential votes, Silver offers insight into how views of the American electorate change over weeks and months.