To use polling data wisely, it's important focus on general trends rather than any single poll. And always remember—election outcomes are unpredictable, and surprises are always possible.
14th October 2024
Digital Worldwide News
As we inch closer to the 2024 US presidential election, one thing is clear: polls are everywhere. The media, both big and small, frequently fall into the trap of "horse race" journalism. Polls dominate the headlines, with reports flip-flopping between Kamala Harris and Donald Trump. One week Harris has the upper hand, and the next, Trump has an edge.
But the real question is—can we trust these polls?
In the United States, presidents aren’t chosen by direct popular vote. Instead, they’re elected through the Electoral College, a system that gives each state a set number of electors based on population and Senate representation. This unique setup means that smaller states wield disproportionate influence, making national polls less relevant.
The "winner-takes-all" system used by most states further complicates polling accuracy. Candidates who win the popular vote in a state receive all its electoral votes (with exceptions in Maine and Nebraska). The most important surveys, therefore, are those conducted in battleground or swing states, where neither party holds a dominant advantage.
In 2024, approximately 10 states are expected to determine the election's outcome. Historically, swing states have shown razor-thin victory margins, often less than 1%. These key states—Nevada, Arizona, Wisconsin, Michigan, Pennsylvania, North Carolina, and Georgia—hold a combined 91 electoral votes. Both Trump and Harris need 270 electoral votes to win the presidency, making these swing states the ultimate deciding factor.
But polling in swing states comes with challenges. When the margins are so tight, even a slight polling error can drastically affect predictions.
In 2016, national polls correctly predicted Hillary Clinton's popular vote victory. She won nearly 3 million more votes than Trump. However, polls missed the mark in key swing states, allowing Trump to secure an Electoral College victory.
Several factors contributed to this miscalculation. According to the American Association for Public Opinion Research (AAPOR), polls underrepresented Republican voters and overrepresented college-educated Democrats. Moreover, undecided voters were underestimated, many of whom ended up casting their ballots for Trump or third-party candidates.
Despite efforts to correct these issues, polling errors persisted in 2020. This time, the Covid-19 pandemic complicated the task. Pollsters underestimated Trump's support in swing states and overestimated Biden's national lead, making 2020 the least accurate election polling in four decades
Despite these inaccuracies, Joe Biden won the 2020 election, securing 4% more of the popular vote than Trump. He also took home 306 electoral votes compared to Trump's 232. Key victories in swing states such as Arizona, Georgia, Michigan, Nevada, Pennsylvania, and Wisconsin made the difference.
These polling errors, however, have had lasting consequences, contributing to increased public distrust of polls, particularly among Republicans.
Polls naturally face skepticism, especially when they repeatedly get it wrong. Republicans, already wary of establishment institutions, are less inclined to trust mainstream polls. In 2016 and 2020, Trump voters didn’t hesitate to express their preferences but were often underrepresented in polls due to distrust of traditional media.
This underrepresentation has led to a recurring issue in polling: working-class white voters’ opinions are frequently missing from the data. Moreover, polling technology faces challenges, such as lower response rates. Reaching voters by phone now requires hundreds of calls, thanks to caller ID and call screening.
To combat low response rates, many polling firms now use a mix of methods, including email, online surveys, and robocalls. While cheaper, online surveys often attract voluntary participants who may not represent the larger population. The increase in online polling has led to a doubling of polling companies between 2000 and 2022, according to Pew Research Center.
But while technology has evolved, the problems remain. Smaller polling samples (fewer than 1,000 respondents) can skew results, and the margin of error is often higher for certain demographic groups.
Another challenge is identifying "likely" voters. Only about two-thirds of eligible Americans actually vote. In 2016, turnout among Democrats was overestimated, giving the false impression that Clinton was a guaranteed winner. Polls suggested Trump was behind, but his supporters turned out in full force while some Clinton voters stayed home.
Understanding the margin of error is also crucial. It’s typically 3 to 4 percentage points but can be higher for smaller demographic groups like young people or specific ethnic groups. Media headlines often exaggerate poll results, claiming a candidate is ahead when the difference falls within the margin of error.
Researchers at the University of California, Berkeley suggest that to ensure 95% accuracy, the margin of error should be closer to 6%. This means a candidate projected to receive 54% of the vote might actually land anywhere between 48% and 60%.
The 2022 midterm elections offered some hope for polling accuracy. These polls were the most accurate since 1998, showing no significant bias toward either party. Still, midterms are different from presidential elections, and the dynamics of 2024 may prove more unpredictable.
That said, many polling firms have adapted their methods since 2016. By 2022, 61% of polling companies had adjusted their sampling techniques, while more than a third made changes after the 2020 election.
Despite the challenges, polls remain useful for gauging general trends and public opinion on key issues. However, they’re far from perfect. Polling methods vary across firms, introducing biases that can make comparisons difficult.
Polling aggregators, like FiveThirtyEight, offer averages that may be more reliable than individual polls. But even aggregators have their uncertainties. After ABC acquired FiveThirtyEight in 2023, its founder, Nate Silver, left to start a new platform, Silver Bulletin, which continues to attract media attention for its forecasts.
With the unpredictability of polls, political betting markets have emerged as an alternative. Platforms like Polymarket have surged in popularity, with some, like Elon Musk, claiming markets provide better forecasts than polls. However, this claim is yet to be proven, and there are concerns about potential market manipulation to sway public opinion.
While polls can provide a snapshot of public opinion, they’re far from reliable indicators of election outcomes. In the 2024 election—potentially one of the closest in recent history—polls should be taken with a grain of salt.
For example, the wording of poll questions can influence results. In 2019, USA Today ran a headline that read: "Poll: Half of Americans say Trump is victim of a 'witch hunt' as trust in Mueller erodes." The poll question combined two different ideas—whether the investigation was a “witch hunt” and whether Trump had been unfairly targeted. This led to biased results, which Trump used to his advantage.
Polls can still be useful, but interpreting them requires caution. It’s essential to look at the fine print—sample sizes, polling dates, margin of error, and methodology all play a role in accuracy.