COMMENTARY
I've previously delved into reading the early tea leaves about the November 2024 U.S. election, but I left out one crucial area: polling. I did so because March was way too early to be trying to make predictions based on polling. Even though we’re now a few months down the line, I hate to break it to you, and a lot of folks, but it’s still too early to be paying attention to polling! For your own sanity, I recommend you stop paying attention to polls until after Labor Day once Election Season (the nerdiest of all the sports seasons), and particularly the post-season kicks off. To put it in sports terms, traditionally the end of the major party conventions is Opening Day, post-season starts after Labor Day, and the championship round start in October. You don't have to take my word for it. Let’s look at polling data since 1980.
In six of the past 11 elections, early summer polling either got the ultimate winner “wrong”, though I’m not a huge fan of that language as opinion likely genuinely shifted, or showed a tie. Basically, early polling is about as accurate as a weather forecast from a groundhog when it comes to reflecting the winner. By September, this number drops to four out of 11. Once we reach October, only three out of 11 polls either got the winner “wrong” or showed a tie. In the two ties, the eventual winner caught up to the early “leader” by October. Further the two late polling ties proceeded election results with the popular vote and electoral college splitting, suggesting close elections make for difficult polling (and polls that have results within the margin of error are not “wrong”). Basically, polls in October do a better job of reflecting the eventual winner, but are not perfect as polls are not forecasts.
So, why might polling be less useful now than in four months? For starters, depending on the poll, about 10% to 20% of voters are still undecided. That's a significant portion of the electorate still deciding! Especially considering this election will likely come down to a few percentage points in a handful of states. I’d like to see more polling firms ask follow-up questions to undecided voters about which direction they’re leaning, especially as the election approaches. It may be that undecided voters strongly overlap with those who dislike both Presidents Biden and Trump. The critical question then becomes who they dislike more. As I’ve written about elsewhere, strength of dislike is a powerful motivator of vote choice these days considering voters dislike just about everything and everyone.
Secondly, most current polls target “all adults” or “registered voters.” As we get closer to the election, polls will shift to “likely voters.” These groups are vastly different, which can impact results. Developing what is referred to as a “likely voter model” is quite difficult as the electorate has been in flux since at least 2016, leading to differing models and results across pollsters. 538 offers a helpful tool that allows you to explore the impact of changes in voter demographics on election results. While not perfect, this tool illustrates why predicting the 2024 electorate's composition may matter for polling results, especially as higher turnout elections, which this should be, look different now than in the past.
Another factor is third-party candidates. Recent polling suggests that about one-third of current third-party supporters may shift before election day. Given that third-party candidates collectively poll around 12-15% at the moment, this means at least 4-5% of all voters could change their minds. This 4-5% could have serious implications for the November outcome. Additionally, ballot access shapes third-party candidate votes totals. Robert Kennedy Jr. is currently the strongest third-party candidate, but is currently on the ballot in a limited number of states. However, it seems likely he will be on the ballot in enough states to be eligible to win 270 electoral college votes by November. RFK Jr. may also qualify for at least one of the presidential debates, which could influence things. Only twice before, with John Anderson in 1980 and Ross Perot in 1992, has a third-party candidate shared the debate stage with major party candidates. Given all the uncertainty around third-party candidates, particularly RFK Jr., it’s possible support for them may shift, especially with key groups, and those shifts could really matter for the final outcome. Plus, prominent third-party candidates may increase, and change the demographics of, voter turnout.
Further, an often under considered factor in early polling is also something called “differential nonresponse”. This is somewhat related to the “silent Trump voter” hypothesis. Basically, this means that supporters of one candidate may be less likely to participate in polling at different points in the election cycle, but particularly before the major party conventions. As we get closer to the election, more supporters may take part in polling. However, this may be shaped by bad news for candidates, like a difficult foreign policy situation that is unpopular with key voting blocks (*cough*) or a criminal trial (*cough*). As researchers put it "when things are going badly for a candidate, their supporters tend to stop participating in polls." This matters more now than it used to because response rates to surveys have tanked and so changes, even slight ones, in who is willing to take part in a survey have a bigger impact on polling during close elections.
Finally, remember that polls are not elections, and we're still five months out. A lot can change in that time. There is still time for an “October Surprise” (along with a June, July, August, September, and early November one). Considering the impact of major events on recent presidential elections, including the late October 2016 announcement from the FBI regarding Secretary Clinton’s emails and the COVID-19 pandemic in 2020, a lot can happen between now and November that may mean polling now does not match the result later this year.
In summary, using current polling to predict the election is like predicting a sports championship before the season starts. We’re still in the pre-season of the November 2024 election, so no matter who is “up” in the polls no one is “losing” or “winning”, no matter what your favorite cable news talking head is saying. Regardless of which campaign you support, don’t overreact to current polls. 2024 is most likely going to be a close one and surveys are allowed to have margins of errors, with several key states likely to have results reasonably within those margins of errors (which will probably be ignored by some prominently loud voices post-election). Plus, keep in mind that the further we are from the election, the higher the uncertainty in polling results. Otherwise, we’d be teaching about the Ross Perot administration in high school history. I must have missed that day.
