A:

Polls got a bad rap after the 2016 and 2020 presidential elections because they were so far off the actual results. Just to remind you, three weeks before the 2016 election, the New York Times pollster Nate Cohn boldly proclaimed that Hillary Clinton’s chances of winning were 91 percent. 

That was a big miss.

The 2020 polls were even worse, with pre-election polling off by the largest magnitude in 40 years, according to Pew Research. It found that polls overestimated support for Joe Biden by nearly 4 percentage points in the national popular vote, the most significant polling error since 1980.

So what happened? 

Mary Snow, an assistant director of the Quinnipiac University Poll, says the first thing to note about 2016 is that the 91 percent chance of winning the election was a forecast, not an actual poll. It’s easy to confuse the two. To be fair, Clinton did win the popular vote by 2.9 million, but she lost the electoral college when Donald Trump won the Blue Wall states: Pennsylvania, Michigan and Wisconsin. One analysis found that if just 40,000 votes in those states had shifted to Clinton, she would have become the 45th president. 

The problems with the 2020 polling are harder to tease out. The American Association for Public Opinion Research or AAPOR says Trump’s voters were the least likely of all groups to participate in polls, a phenomenon sometimes described as the “shy Trump effect.” It’s also possible that the polls did not capture voters who rarely vote but showed up to support Trump. Pollsters call this “partisan nonresponse bias.”  The result is that rural white, working-class, and non-college-educated voters were underrepresented, according to Republican pollster Patrick Ruffini.

To ensure history does not repeat itself, pollsters are adjusting methodologies and trying new strategies in 2024. According to Snow, internal research at polling organizations has resulted in new methods to reach voters. The gold standard used to be phone calls, but as you can see in the chart below, some pollsters now use a combination of online, mail, phone outreach and probability-based panels, which are a list of residential addresses drawn randomly from the U.S. Postal Service’s database.

There’s also a change in how pollsters correct for any differences in their sample survey and the actual US population. Previously, polls were weighted to represent US demographics on variables such as age, race and gender. However, pollsters like Pew Research and The New York Times/Siena College now adjust their data for 12 variables, such as education, party identification, political engagement and voter registration status to get more accurate results. That seems to be working for The New York Times/Siena poll, which is the highest rated pollster, according to 538.com.  Experts say these top 10 are among the best pollsters in the country because they are independent polling organizations with no vested interest in the results. You can also compare 538.com’s list with the New York Times poll tracker where a diamond signifies a higher quality poll.

So, if you want to know where the race for the White House stands, check out the highest-rated pollsters that use a multi-modal approach to reach voters, and then look at the sample size. A sample size of 1000 people for a national poll and 600-700 voters for state polls is considered more reliable because the margin of error will be smaller. 

But even with smaller margins of error, like plus or minus 3 percentage points as you see in the shaded lines below, the race is really a statistical dead heat unless one candidate is winning by double the margin of error. So in this case, Harris would need to be at 52 percent to be confident that her lead is not just a sampling error. 

Another caveat: No one poll is going to provide 100 percent certainty. That’s why Snow prefers to look at polling averages like this one by Nate Silver at Silver Bulletin. “A single poll could be off,” says Snow. “As a news consumer, don’t put too much credence in one poll. Look at trends and averages.”

Polls are also not adept at providing voters with information and context about the big election issues behind those statistics. Instead, they turn politics into “a savvy analysis of who was up, who was down, who’s winning or likely to win, the horse race, the spin, the strategy,” according to media critic Jay Rosen, who wants journalists to focus on “not the odds, but the stakes” in this election.

In the end, the big wild card in every election is turnout–especially in the key swing states. Pollsters make an educated guess about who will actually cast a ballot, but it’s enormously difficult to predict the actual makeup of the electorate. “Most people have their minds made up, so it’s all going to be about turnout,” says Snow about the 2024 election. “That’s what will determine this race.” 

So remember, polls are not crystal balls—they are only a snapshot of public opinion at a single point in time. “Don’t expect polls to have the exactly same result as the election,” says Snow. “And if a pollster says the election is too close to call, believe them.”