By Dr Anandadeep Mandal
Birmingham Business School, University of Birmingham
The reliability of 2024 U.S. election polls is a critical question given recent polling inaccuracies in key election years. Historically, polls have struggled to capture precise snapshots of voter preferences, especially when predicting voter turnout, demographic preferences, and unexpected political shifts. Polling accuracy depends on several factors, including sampling methodologies, response rates, and models predicting voter turnout. In the 2016 and 2020 elections, these factors contributed to significant errors, particularly underestimating Republican support and overestimating Democratic leads. Understanding these challenges, along with the improvements pollsters are making, can help us assess how much we can trust polls in 2024.
Historical Polling Errors: 2016 and 2020
Polls performed poorly in 2016 and 2020 due to several compounding factors. In 2016, many polls underestimated support for Donald Trump, especially in swing states. Nationally, Hillary Clinton led by an average of 3% in final polls but won the popular vote by just 2.1%, while several key state polls fell short, predicting her victory in states like Pennsylvania and Michigan, which Trump won.
In 2020, these errors were even larger. Polls suggested Joe Biden would win by a margin of 8-10% nationally, but he secured a 4.5% lead in the popular vote. A 2021 review by the American Association for Public Opinion Research (AAPOR) found that this 4.5% polling error was the highest in 40 years for national polling. In battleground states, polling errors reached over 6%, primarily due to under-sampling Trump supporters. This discrepancy has been attributed to multiple factors, including “non-response bias,” where certain demographics, particularly Republican-leaning voters, were less likely to participate in polls.
The “Shy Trump” Effect and Sampling Bias
The “shy Trump” hypothesis posits that some voters, reluctant to disclose their true preferences, might underreport support for Trump. However, studies conducted by the Pew Research Centre and AAPOR show little to no evidence of this effect being significant. Instead, errors in 2016 and 2020 were largely due to other forms of bias. For instance, some Trump voters, particularly non-college-educated white voters, were less likely to respond to polls. This “non-response bias” skewed results and is particularly challenging for pollsters, as these groups are harder to reach and are often underrepresented in polling samples.
Voter Turnout Models: A Key Source of Uncertainty
Accurately predicting who will vote is one of the most significant challenges for pollsters. Turnout models are typically based on past voting patterns, but they rely heavily on assumptions about demographic behaviours that may no longer apply. For instance, while Trump drew significant support from irregular voters in both 2016 and 2020—voters who typically abstain from midterms but show up for presidential elections—these unpredictable patterns complicated turnout forecasts. Pollsters often rely on “likely voter” models, which can miss irregular patterns or shifts in political engagement driven by current events, social movements, or last-minute campaign strategies.
Improvements and Challenges in 2024 Polling
Pollsters have been actively working to address issues that impacted previous elections. Methodological adjustments include broadening survey outreach to account for harder-to-reach demographics and increasing online and text-based polling methods, which tend to have higher response rates among younger and more diverse voters. This shift has improved participation but also introduces its own challenges. For instance, online and text-based polls must ensure balanced samples to avoid over-representing certain demographics, such as tech-savvy younger voters.
Moreover, recent polling, like that during the 2022 midterms, has shown improved accuracy, with error rates under 2%. This accuracy increase has encouraged some confidence in the polling community; however, the stakes of presidential elections, combined with turnout unpredictability, create a more complex environment for pollsters. As turnout models are still based on educated guesses, they remain vulnerable to shifts in public sentiment, breaking news, and major events close to election day.
How Polling Data Should Be Interpreted
While the improvements in polling methodologies are promising, interpreting polling data accurately requires understanding its limitations. Polls generally provide a snapshot of public sentiment at a specific point in time but are less reliable at predicting narrow races, especially in battleground states where margins can be razor thin. In tight elections, even small errors—between 2% and 4%—can change predictions significantly. This potential swing makes it essential to view polling data in 2024 as an indicator of broader trends rather than a definitive predictor.
Pollsters also emphasize that polling data remains valuable for gauging issues of public concern, such as the economy, healthcare, or climate change. By analysing which issues resonate with specific demographics, campaigns can focus their efforts, regardless of the accuracy of “horse race” metrics. Thus, polls serve more to identify trends and gauge public opinion on issues than to declare an outright winner, especially in closely contested elections.
Can We Trust 2024 Election Polls?
With new methodologies and lessons from past mistakes, polls in 2024 may be more reliable than in recent elections. However, historical polling errors reveal that even small miscalculations in turnout models and demographic responses can lead to inaccurate projections in close races. Therefore, while polling data provides valuable insights into voter sentiment and issue preferences, the public should approach 2024 polls with caution. The 2022 midterms showed that poll accuracy can improve, but predicting turnout patterns, especially with a high-profile figure like Donald Trump possibly on the ballot, is inherently uncertain.
Ultimately, while polling data remains a powerful tool, its real strength lies in showing trends, rather than predicting outcomes precisely in close races. Election watchers, therefore, should use polls to gain insight into public sentiment and the broader electoral landscape, recognizing their potential margin of error and focusing on aggregate trends to avoid overestimating individual polls. By interpreting polls as directional indicators, we gain a more nuanced understanding of the likely dynamics of the 2024 race while remaining realistic about their limitations.
-
- Read more about Dr Anandadeep Mandal
- Back to Social Sciences Birmingham
The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of the University of Birmingham.