Has The Country Lost Trust in Polls?

Especially during this most recent presidential election, many wonder about the accuracy of polls and how they are conducted. Photo by Matthew Hurst via Flickr

The results of the 2024 presidential election are in: Donald Trump is set to become the 47th President of the U.S. Despite polls predicting a close race — with Vice President Kamala Harris ahead by 1.2% — the election ended in a sweeping victory for Republicans. Trump won 312 electoral votes, well over the 270 needed. This is not the first time polls have incorrectly predicted the winning candidate. 

A 2024 Marist Poll survey found that 61% of Americans have not very much or no confidence at all in polls. In 2016, national polls generally predicted a close race with a slight edge for the Democratic candidate, Hillary Clinton, who ultimately won the popular vote. However, Trump secured the presidency by winning 304 electoral votes, surprising many analysts.

In 2020, polls favored President Joe Biden, who won with 306 electoral votes to Trump’s 232, but some polls had suggested a closer race in key battleground states. These discrepancies contributed to ongoing skepticism about polling accuracy.  

However, Lee Miringoff, Director of the Marist Institute for Public Opinion, has a different theory. 

“The polling we did in 2016 was actually very accurate…the problem was not as much in the pollsters, but in the forecasting models,” said Miringoff. 

He holds that these forecasting models, which are based on polls, gave people a false perception of who was likely to win the elections. Forecast models, also known as regression models, are based on prior elections to help determine the votes for each political party in current elections. 

Polls are generally taken by a limited sample of respondents, and their answers are then applied to the larger population to gauge public opinions. Random sampling is a common strategy used by surveyors to reduce the probability of bias in their statistics. The results are then interpreted using a few variables, such as age, race and gender. A margin of error is determined to establish how much the results of a poll may differ from the true population value. The higher the margin of error, the less confident people can be in the poll. 

Several factors can affect the accuracy of a poll. For example, non-response bias is a major issue with polling, which occurs when the people contacted do not respond and therefore are not represented in the poll. Social desirability bias is also a common issue, when respondents may answer a poll based on what they think is most socially acceptable, instead of what they agree with. This leads to an inaccurate representation of the respondents’ true values. 

The Marist Poll, a nationally recognized public opinion poll operating on the Marist College campus, has provided accurate and reliable information since 1978. It was the first college-based research center that included undergraduate students in conducting its research and has adjusted its polling techniques in recent years to gain a more accurate representation of the population.

“We have to make sure we’re using multiple other modes of collecting data, including text-to-web for example, where we reach out to people, and then, they go to a link and fill out the survey,” explained Miringoff. “But then again, you’re missing people who don’t have access to online answers.” 

Miringoff also believes that another important factor of a poll’s credibility is its transparency. 

“I think it’s important for both an educational purpose for our students, but also for the public, that we’re very transparent in our methods,” explained Miringoff. 

The Marist Poll publishes a detailed explanation of how each poll is conducted, as well as the demographics of the people surveyed for every poll. 

Pollsters have been adjusting their methods in recent years, including increasing the number of active polling organizations, which allows for more diversity in the polls. According to the Pew Research Center, the number of organizations that conducted national election polls more than doubled between the years 2000 and 2022. 

The mediums used for polling have also changed over the years. In previous years, most polls were conducted by telephone. Today, pollsters have begun using mostly online methods, or a combination of mail, telephone and online methods. Doing so has allowed polls to reach a more diverse set of demographics, particularly younger people who are less likely to answer phone calls.  

In the end, polling remains a tool of estimation, not prediction. It is essential for both the public and pollsters to understand the limitations of these surveys. As Miringoff explained, “It’s important to know that, when you see a poll number, we’re not talking about the exact precise number, but a range within which the response would fall had we interviewed everybody.” 

“All polling is a snapshot. It’s an estimate within a range,” said Miringoff.