Do Early Polls Accurately Predict Election Results?

Recent polls indicate that Trump maintains a slight lead in the RealClearPolitics Polling Average. However, looking at polls for an election that is still five months out raises the question: Do early polls reliably predict the final election results? The short answer: They do – at least in recent U.S. presidential elections.

Looking at data from the past five presidential elections, every candidate who polled better in the national popular vote at the start of the election year won the national popular vote in November. There was never a change mid-election season where the person behind in polling became the frontrunner and won the national popular vote. One caveat, however: The national popular vote doesn’t always produce the winner. In 2016, when former Hillary Clinton won the popular vote by nearly 2.9 million votes, Donald Trump won the presidency. (The margin was much closer, but the same thing happened to Al Gore in 2000.)

Based on data from the last five elections, Trump would be the predicted victor of the popular vote this year, as polls have consistently shown him ahead since the start of the year.

In addition, it's important to note that if one goes back further in time, the track record for frontrunners isn’t perfect. Christopher Wlezien, a University of Texas at Austin professor of government and co-author of the book “The Timeline of Presidential Elections: How Campaigns Do (and Do Not) Matter,” has delved into the topic of early polls. His research reveals that while early polls do sometimes predict the eventual results, they are not infallible predictors. 

Along with co-author Robert Erikson, Wlezien examined presidential elections since 1952 to see how polls change throughout the election year. In the summary of their findings, they write, “Polls from the beginning of the year have virtually no predictive power.” Their explanation is that in most elections, the candidates aren’t fully known until after primary season, so voter preferences are still variable.

Yet Wlezien found that polls become more accurate after the primary season. On June 7, Wlezien told RealClearPolitics in an interview that in five out of the 18 election years since 1952, the leader in the polls a few weeks before the conventions went on to lose. 

To accurately predict election winners, Wlezien and Erikson wrote in their 2020 paper "Forecasting the 2020 Presidential Election: Leading Economic Indicators, Polls, and the Vote," prognosticators must combine post-convention polling data with economic indicators. The model they developed using this method would have predicted 15 of the last 18 presidential elections correctly. 

One of their main findings was that the turning point in polls, where polls went from fairly accurate to almost always accurate, was the party conventions. Through campaigns and the conventions, they write, "voters are made aware of—or not made aware of—fundamental factors like candidates’ policy positions that determine which ticket will get their votes." Wlezien emphasized this point, telling RCP that a large part of forming the swing voter’s opinion comes from the conventions: “That's part of what the conventions do. They help those folks. They help bring home those fundamentals.”

This increase in voter knowledge in post-convention polls is more accurate as people have greater knowledge of the candidates and the issues they are campaigning on. According to Wlezien, every single election since 1952 has been accurately predicted by polls around two weeks after the last convention. "By the time we come out to the convention, post-Labor Day, the leader of the popular vote [in the polls] has always won" the popular vote in the November election, he told RCP. 

With the Republican National Convention in Milwaukee July 15-18, the Democratic National Convention in Chicago August 19-22, and the first presidential debate later this month in Atlanta, Trump’s lead in the polls could still change before voters’ opinions are solidified.