### Polling Concerns Arise as Election Day Nears: Experts Warn of "Herding" and Reliability Issues

As the 2024 election draws closer, experts in political polling are expressing concerns about the accuracy and reliability of the polls, pointing to a phenomenon known as "herding." Herding occurs when pollsters, wary of damaging their reputations, tend to align their results with what is widely considered the likely outcome rather than risk presenting outliers that could be entirely wrong.

Polling is a complex science, involving dozens of firms that survey thousands of people through various methods such as mail, phone, and online surveys. Pollsters often combine these approaches and implement techniques like weighting responses to improve accuracy. Despite these efforts, certain patterns and biases can still impact the results. One notable example is the tendency to present similar outcomes, particularly in the final days of an election, which polling expert Nate Silver highlighted on Oct. 29. Silver raised alarms on X (formerly Twitter) about the lack of variance in swing state polls, which consistently show a tight race between Trump and Harris.

Outliers do exist, however. The Times/Siena College poll, for instance, has a history of publishing predictions that deviate from the norm. Although it predicted a nine-point win for Biden in 2020 when the actual margin was 4.5 points, it remains an exception to the herding trend. Silver also noted a Des Moines Register/Mediacom poll showing Harris with a three-point lead in Iowa, contrary to other projections favoring Trump.

Another challenge is the timing of the polls. Most do not capture data from Election Day itself. For example, the Times/Siena poll publishes its final national results towards the end of October, with data collected by October 23. Given that early voting begins weeks before the actual election day in many states, any last-minute shifts in voter sentiment go unrecorded. While significant developments in the days preceding the election rarely swing the outcome, this temporal error remains a concern.

Polling data is inherently flawed, as explained by FiveThirtyEight. Even with perfect polling, the results would likely miss the final margin by roughly two points. This margin of error, typically two to five points among many polls, becomes crucial in a tightly contested election like 2024. Recency and nonresponse biases also underline the limitations of current polling. The former bias suggests that voters might expect Trump to win because of his 2016 victory, while the latter points to the habit of certain groups, notably Trump supporters, being less responsive to surveys.

Weighting is another practice pollsters use to address underrepresented demographics, but it carries risks. For the 2018 midterms, weighting by education helped correct the 2016 inaccuracies. Recently, pollsters have been weighting based on "recalled vote," where respondents report their past voting behavior. This approach aims to ensure enough Republicans are reflected but can skew results if memories are faulty. Silver has criticized this method as another form of herding, aligning current polls too closely with past data.

The "Shy Voter Theory" suggests Trump supporters might not always be forthright about their voting intentions, potentially leading to underestimates of his support. However, Silver contends many Trump voters are now more open about their preferences, possibly reducing the impact of this factor. Another theory, the "Bradley Effect," speculates that some white voters may not admit to not voting for a Black candidate, a factor that might affect Harris as the first Black woman presidential candidate.

Various theories attempt to explain the discrepancies in polling accuracy, with the "Unified Theory" attributing it to less engaged Trump voters who are less likely to respond to surveys. In contrast, the "Patchwork Theory" attributes inaccuracies to unique factors influencing each election.

Concerns about the 2024 election polling largely stem from the inaccuracies observed in the 2016 and 2020 elections. Trump's passionate, often less-engaged voter base has historically been underrepresented in polls, leading experts like Nate Cohn to suggest that current polls might be either underestimating Harris or overestimating Trump's support. As the tight race continues, the accuracy of these polls remains under scrutiny.