Pollsters struggle to get Trump supporters to talk to them while Trump haters cannot wait to talk. The resulting Republican samples are wildly off track.
The latest New York Times poll shows Vice President Kamala Harris with a slim three-point lead over former President Donald Trump, but the result is well within the margin of error. A closer look at the data, however, prompts a pressing question: why are so many polls getting it wrong?
Polling is meant to capture a snapshot of public opinion, yet the New York Times survey of 3,385 likely voters—conducted over eight days—fails to meet that standard. With events like a hurricane and a vice-presidential debate occurring during the polling period, daily fluctuations in responses seem inevitable. But the real issue lies in who was asked for their opinions.
A key problem with this poll is its lopsided sample. According to the data, 33.3% of respondents identified as Democrats, while only 26.5% identified as Republicans. National voter data, however, shows that Republicans now outnumber Democrats by about 3%. Why then, would the New York Times poll sample contain 7% more Democrats than Republicans?
This discrepancy matters. Despite the poll’s slight lead for Harris, it still shows Trump outperforming with independent voters. If Republicans had been properly represented, that three-point lead could easily shift in Trump’s favor.
The poll suggests only 89% of Republicans back Trump, a number that’s hard to believe given his dominance within the party. Pollsters are struggling to find Republican voters—especially Trump supporters—willing to participate in their surveys. Those who are willing to talk to pollsters are anti-Trump Republicans, skewing the GOP sample by overrepresenting Trump haters, further tilting the results to Harris.
To compensate for this imbalance, many polls use "weighing," multiplying Republican responses to make up for the lack of participants. But here’s the problem: if you're working with a flawed sample, multiplying it just amplifies the errors. In effect, the anti-Trump voices get multiplied, and the poll results shift even further away from reality.
This issue isn't new. In 2016, many polls underestimated Republican turnout, leading to widespread shock when Trump won key battleground states like Pennsylvania. Since then, Republican voter registration has surged. In Pennsylvania alone, the gap between Democrat and Republican registrations has shrunk from nearly a million in 2016 to just 300,000 today. Yet the New York Times sample suggests that Republicans will make up only 26.5% of voters, a clear misrepresentation of current trends.
So, why are the polls wrong? It comes down to faulty sampling, compounded by efforts to adjust flawed data. As voter demographics shift and pollsters struggle to reach Trump supporters, these errors are likely to continue. If post-election results differ drastically from what the polls suggest, this will be the key reason why.
.png)