Election Crystal Balls: Are Presidential Polls Accurate?

Did you receive a call or text about participating in a poll during this past election? Did you respond? Perhaps you were skeptical or felt they lacked validity. Recently, polling organizations have received criticism because presidential election predictions seem far from actual results. In the 2020 election, the polls reported the largest errors seen in 40 years, leading many people to distrust them. However, polling is deeply woven into the history of United States elections. Contrary to their use today, the first polls were never even intended to predict election outcomes.

 

The History of Polling the United States

In 1824, the first-ever recorded presidential polls took place. These polls started out informal and a bit chaotic. Voters conducted them anywhere from grand juries to Fourth of July celebrations. They used straw polls– no, not literally picking straws– but headcounts of people’s preferences to gauge each other’s opinions. 

The first modern statistical poll was conducted in 1936 by George Gallup. At the time, Gallup challenged popular polling methods and created his own survey using quota sampling. This method attracted and randomly surveyed various groups of eligible voters across the country. Gallup’s polls included a broader range of respondents, such as low-income voters and women, making his results more representative, unlike his competitors. His poll correctly predicted the winner, President Franklin D. Roosevelt, while other poll predictions were incorrect.

Are Polls Accurate?

Presidential polls have never been 100% right all the time. In fact, one major mishap occurred in 1948 when Gallup and the Chicago Daily Tribune mistakenly declared “Dewey Defeats Truman,” immortalizing the famous photo of a reporter triumphantly holding the incorrect headline. Research suggests that polling data has not gotten more inaccurate or biased in recent years. Moreover, it is becoming the opposite as pollsters* adjust methods to solve previous survey mistakes. 

The Challenges Faced by Pollsters 

Pollsters face a major challenge: the unpredictability of voters like you! Just kidding–sort of. Survey data often falls short in predicting people’s behaviors, leading to incorrect predictions about how they vote. If an unexpected event were to shift public opinion about a candidate, or if there were a significant change in voter turnout on election day, the polls could not account for that.

Traditionally, polls have used probability sampling, which randomly selects a group of people within a population. Polls cannot survey all eligible voters, but everyone should have an equal chance of being surveyed. Usually conducted through random digit dialing(RDD), this method allows researchers to select telephone numbers to survey voters randomly. However, response rates for telephone surveys have dropped significantly since the 90s. With the rise of automated telemarketing calls to cell phones, people are reluctant to pick up “spam risk” or “no-caller ID” calls. I sure am! 

Pollsters have opted for more non-probability sampling strategies because of the need for more participation. One example is opt-in surveys, where everyone visiting a website is invited to complete a survey. However, a consequence is that certain people are more likely to take surveys than others. Studies show that older and college graduates are more willing to take surveys and are overrepresented in polls. This can skew results in favor of candidates these groups are more likely to support. 

To fix this, pollsters must weigh critical factors related to people’s behaviors and characteristics when analyzing their data. Factors such as age, income, political affiliation, and voter turnout can be considered to make predictions closer to reality.

Previous Polls & Elections

In 2016, these factors caused problems. When voters in key states decided on their choice in the election’s final days, polls did not have data on who they would vote for. The polls failed to account for the number of undecided voters*, leading to biased results. Another issue was that people with higher education were more likely to participate in the polls and preferred one candidate over the other.

Noting the mistakes from the past election, popular polling organizations adjusted their methods in 2020. They included the likelihood of undecided voters and accounted for specific groups’ overrepresentation in surveys. Yet, most polls incorrectly overstated the electoral votes for Biden over Trump. Why? While it is less clear in 2020, pollsters have a few theories. 

First, pollsters predicted that the number of voters belonging to each party was the same as in 2016. In reality, they underestimated the number of Republican voters and overestimated the total number of voters that turned out. Analysts also believe that some voters withheld their true preferences from polls by not participating or changing their vote choice. 

Though there were still errors, the polls were more accurate in this election. Even as votes are still being counted, pollsters suggest that, again, they underestimated the number of Republican voters. 

What Can You Do With the Polls?

In the next election, you may find it beneficial to contemplate what the polls mean to you. If you find yourself checking the polls daily or feeling anxious about them, it’s okay to take a step back. While it would be incredible if polls were like a crystal ball magically predicting the next president, that is not the entire reality. Polls cannot always correctly forecast, but they are a profound part of U.S. elections.

*Pollster: a person who conducts or analyzes opinion polls from surveys about individuals’ preferences. 

*Undecided voters: individuals who are likely to vote but did not choose who to vote for before election day.

 

References

1. https://news.vanderbilt.edu/2021/07/19/pre-election-polls-in-2020-had-the-largest-errors-in-40-years/
2. https://sites.duke.edu/hillygus/files/2014/06/HillygusPOQpolling.pdf
3. https://www.pbs.org/fmc/segments/progseg7.htm
4. https://journals.sagepub.com/doi/10.1177/00027642221118901
5. https://www.pewresearch.org/methods/2023/04/19/how-public-polling-has-changed-in-the-21st-century/#:~:text=A%20new%20Pew%20Research%20Center,what%20they%20used%20in%202016.
6. https://aapor.org/wp-content/uploads/2022/12/Sampling-Methods-for-Political-Polling-508.pdf
7. https://www.pewresearch.org/short-reads/2019/02/27/response-rates-in-telephone-surveys-have-resumed-their-decline/
8. https://aapor.org/wp-content/uploads/2022/11/AAPOR-2016-Election-Polling-Report.pdf