Missed the Mark: Why Pollsters Misjudged the 2024 Election Part 1

Most political pollsters were wrong about how close the 2024 U.S. presidential election would be, especially in so-called battleground states, which were expected to be toss-ups just days before November 5. Why did the polls underestimate Donald Trump’s popular vote margin wildly and, by extension, his dominance of the Electoral College?

One answer is that public opinion polls, and surveys in general, are seriously flawed. Respondents to questions about their preferences for political candidates or policy outcomes are not asked to make real choices over real options. The cost of stating that candidate A hypothetically is preferred to candidate B essentially is zero. Nothing is sacrificed when the respondent opines one way rather than another or says that he or she is undecided.

We also know that survey results depend on the precise wording of questions. Some are “leading,” biased, or vague. Questioners’ body language, facial expressions, or vocal tones may influence respondents’ beliefs about acceptable answers during in-person or telephone surveys. Opinion surveys administered online raise challenges relating to sample sizes and participant self-selection.

Polling failed miserably in the run-up to Election Day 2024. Pollsters and political pundits consistently misjudged the “red wave” that swept the nation on November 5, missing the mark in the opposite direction from 2022’s midterm elections when broad Republican gains had been anticipated.

Although political candidates and parties closely monitor pre-election polls, they are misleading at best and useless in practice. They are no more informative than the “focus groups” and marketing surveys that produced commercial flops like the Edsel or New Coke. 

Even with sample sizes in the thousands rather than the millions, no one response to a survey determines a poll’s overall results. Error rates in most opinion polls typically amount to three percentage points in either direction.

A vote cast in a mass democratic election also is unlikely to be decisive. As of this writing, more than 140 million voters turned out for 2024’s presidential race. Donald Trump outpolled Kamala Harris nationally by about four million popular votes.

It is impossible to say which of those four million voters put Trump “over the top.” The probability of one vote being decisive or pivotal in 2024 was vanishingly small (roughly one in 140 million). 

The late public-choice economist Gordon Tullock relied on such evidence to conclude that voting in mass democratic elections is irrational because the costs of voting (becoming informed about the candidates and issues on the ballot, traveling to and from the polls, or filling out and mailing a ballot) exceed the benefits of doing so for most voters, measured by the nearly zero chance of influencing the election’s outcome. For Tullock, the probability of one voter’s decisiveness on Election Day is smaller than the chance of being struck by lightning on the way to or from polling places or ballot drop boxes.

However, Tullock also launched a research agenda concluding that voting is rational if seen as a low-cost way of expressing political preferences. Because one vote is not consequential, voters can support candidates or policies they would not support otherwise. Many voters would reject a policy that raises their annual tax bills by $10,000 if a single vote determined the outcome. But if the chance of being pivotal is one in 10,000 (0.0001)—much larger than one in 140 million—a vote in favor of the policy raises the voter’s expected tax bill by just $1 (= 0.0001 times $10,000). Even uncharitable people can afford to be charitable under such circumstances, supporting programs and policies they would reject if they faced the full cost of their choices.

Close elections almost always trigger challenges by losers, ballot recounts, and other consequences that introduce uncertainty into voting processes. Such uncertainty makes it more difficult to conclude that one vote will “matter” ahead of time. 

The point is that opinions expressed in response to political polls and decisions made in the voting booth differ substantially from ordinary market decisions, where costs and benefits are more closely aligned. Because they have skin in the game, people’s market choices do reveal their preferences. Thus, it should be no surprise that pollsters claiming to forecast hypothetical political preferences often get it wrong.