What polls got right and wrong about the 2024 presidential election

The polls mostly got it right.

Broadly speaking, the aggregate of the 2024 presidential election polls indicated that it was an extremely close race between former Republican President Donald Trump, now the president-elect, and Democratic Vice President Kamala Harris. The polls indicated the race would come down to seven swing states.

And that’s exactly what happened. With 46 state race calls as of Wednesday afternoon, neither Trump nor Harris has won a single state predicted to be a lock for the other candidate. There were no major upsets yet in the Senate map either.

Trump’s current popular vote margin is 3.5%, but that could easily narrow as more results come in. Populous states like California still have significant amounts of outstanding votes.

See also >> Yahoo News’ live blog of continued election results

The major election modelers looked beyond individual polls — which can vary significantly — to forecast the big picture. As the election approached the closing days, they focused on the swing state polls, averaging them together and applying weights based on factors like recency and historical accuracy.

Those models showed that either candidate could easily win the election. FiveThirtyEight and Nate Silver’s Silver Bulletin each had Harris’s odds at 50%. Split Ticket put her at 53%. The Economist gave her a 56% chance of winning. These were all essentially coin flips.

But just because the final election odds were forecast as 50-50 between Trump and Harris, that doesn’t mean the Electoral College outcome was expected to be a tie. Silver, for instance, pegged the odds of an electoral tie at just 0.3%.

Ahead of the election, G. Elliott Morris, the editorial director of data analytics at ABC News, which publishes FiveThirtyEight, wrote about how either Trump or Harris could win decisively in the Electoral College with a fairly average polling error.

“Based on how much polls have been off in the past, our election model estimates that the average polling error in competitive states this year will be 3.8 points on the margin. This error is not uniform across states … but, generally speaking, when polls overestimate a candidate, they tend to overestimate them across the board,” Morris wrote.

“Given that all seven key swing states are so close, even small polling errors in the same direction can have a big impact on who wins the election. According to the simulations from our model, there is a 60-in-100 chance either candidate wins over 300 Electoral College votes.”

In an era of increasing polarization, in which algorithms aim to serve up social media content that users are predisposed to agree with, it’d be easy to approach the Nov. 5 election expecting an overwhelming victory for your preferred candidate.

For Harris supporters, that often included analysis about why they could expect their candidate to easily sweep most of the swing states. There were individual polls, when isolated, that could support this view.

Ann Selzer’s surprise poll of Iowa, published by the Des Moines Register, provided a sharp glimmer of hope for Harris’s supporters that the polling consensus was wrong. Her highly rated survey showed Harris slightly ahead in Iowa despite the state’s strong Republican lean. That poll ended up being wrong, as Trump is currently holding onto a 13-point lead in the state, but other Iowa pollsters got it much closer, pegging Trump’s winning margin in the high single digits.

And there were various theories about why the polls might be undercounting each side’s preferred candidate. Harris supporters argued that the surveys were overcorrecting after undercounting Trump voters for the past two election cycles. Other people on social media accused Republican pollsters of intentionally skewing the averages with a small flood of Trump-favorable surveys.

Comedy Central host Jon Stewart bluntly summed up much of the frustration. “I do want to very quickly send a quick message to all the pollsters,” he said during his channel’s election night broadcast. “I don’t ever want to f***ing hear from you again. Ever. I don’t ever want to hear, ‘We’ve corrected for the overcorrection with the voters!’ You don’t know s*** about s***, and I don’t care for you.”

The final Yahoo News/YouGov survey had Trump and Harris at 47% of the vote apiece among likely voters. That left 6% of likely voters who were either backing third-party candidates or who were still undecided — giving ample room for either candidate to take a small lead.

Howard University students watch live election results during a watch party near an election night event for Democratic presidential nominee Vice President Kamala Harris at Howard University in Washington, Tuesday, Nov. 5, 2024. (AP Photo/Nathan Howard)
Howard University students watch live election results during a watch party near an election night event for Vice President Kamala Harris. (Nathan Howard/AP)

Polls are almost always at least a bit off. Some of that is due to imperfect assumptions about what the likely voter electorate will look like. Some of that is due to statistical error. In a typical presidential election, the polls have been on average about 4 points off, according to FiveThirtyEight’s analysis going back to 2000.

The biggest misses this year, unsurprisingly, can be found in states where there were the fewest number of high-quality surveys. Some polls had Trump with only a narrow lead in Kansas, for example.

And Trump proved remarkably durable in a number of blue states that Harris was expected to easily win, posting closer-than-expected margins in places like New York and New Jersey. These types of not-so-bad losses will help his national popular vote total.

As of Wednesday afternoon, there are many more votes to count. Control of the House is still up in the air, as are some key Senate races that will help determine by how much Republicans control the chamber next year.

Once all the votes are counted, there may still be some additional polling misses for pollsters and politicos to debate in the weeks, months and years ahead.