As polling becomes more mainstream due to the coming election, many point to the 2016 election and its polling errors to curb Harris enthusiasts who see the election as theirs to win. However, many don’t know that the 2020 election was even further off in favor of Democrats, with states such as Wisconsin having numerous polls well above 10 points only to be 0.63% in favor of Biden once the ballots were counted. With polling errors like these, one can only fear that the same will happen again and the polling is compromised. In this two-part editorial and extension of State of the Race, polling errors will be examined in this article, and then possible errors for this coming election will be examined in part 2.
2016: Breaking the Political Gravity
The errors in 2016 polling have been well documented. Firstly, it is essential to recognize what was said in 2016, namely that Trump was given a nearly 20% chance of winning the election, according to most models. The simple fact is that a 20% chance is nonetheless high enough for that to happen—and it did. Nevertheless, there were many issues with polling.
The first issue was that polling did not fully bake in the James Comey FBI probe into Hillary’s emails. This was her “October surprise,” and the effects of this would take weeks to fully materialize, meaning polls would not show its effects while the ballot box would.
A more pertinent issue is that pollsters did not consider education when weighing results. Due to how polls are done, pollsters must mathematically weigh results to get a more accurate measure of the US as a whole, not just the specific group polled. This weighing traditionally would be around race, gender, age, geographic location, and more. However, education level was not traditionally accounted for, which proved fatal to poll accuracy. Trump was extremely popular amongst a traditionally Democrat voting group: white working-class men. Pollsters did not consider this and did not weigh it by education, meaning that this critical demographic was not accounted for, and the election upset of 2016 occurred.
2020: A Win Casting More Doubt
The issues in polling are often overshadowed by 2016 due to the election upset. Nonetheless, nearly every single major pollster was outside of the margin of error on at least one swing state, if not most (albeit this is not precisely how pollster accuracy is measured). Unfortunately, unlike in 2016, the issues are not as well documented, but two theories have arisen.
The first theory lay in who was being polled. Firstly, as polls soured on him, Trump cast doubt on polls, making his followers less likely to answer polls. Next, Democrats were more likely to stay home during lockdown and thus had more time to answer polls. Thirdly, an unprecedented number of people voted in 2020, meaning pollsters had not accurately weighed the results. Finally, a study was released by NYTimes/Siena that described an issue with some people being polled in that they would state they would be voting for Trump, curse at the pollster, and then end the call, Siena realized that had they counted these voters, 50% of the error would have been resolved.
This creates the second theory: fundamentals. The 2024 538 forecast model methodology describes a concept known as “fundamentals.” Namely unemployment, the stock market, etc. They stated that when accounting for this, their forecast model would have been far more accurate as fundamentals favored Trump even in a Covid environment.
In the end, whether these were the primary causes or not, this does not change the fact that polling was off 3.9% in favor of Biden, creating an extremely tight race.
The question now stands: will there be any issue in 2024?
Part Two Releasing Saturday
Leave a Reply