cover of episode How The Polls Did In 2024

How The Polls Did In 2024

2024/11/21
logo of podcast FiveThirtyEight Politics

FiveThirtyEight Politics

AI Deep Dive AI Insights AI Chapters Transcript
People
E
Elliot Morris
G
Galen Druk
R
Ruth Egelnik
Topics
@Galen Druk :2024年大选对民调机构来说是一次高风险的考验,因为2016年和2020年的民调都出现了重大失误,公众对民调的信任度下降。2024年民调的准确性创历史新高,但仍存在不足,例如连续三年低估特朗普的支持率,以及不同民调方法之间存在差异。我们需要进一步研究以了解民调方法的有效性和局限性,并改进民调方法以提高准确性。 @Ruth Egelnik :2024年民调总体表现良好,但连续三年低估特朗普的支持率(约1-2个百分点)。按以往投票情况进行加权的民调方法似乎有效,但其有效性可能取决于选民群体和具体州的情况。对按以往投票情况进行加权方法应谨慎使用,因为它可能并不适用于所有选举情况。我们需要进一步研究以确定其有效性和适用范围。此外,一些新兴的民调方法,例如“河流抽样”,比传统的在线面板和电话民调更准确。 @Elliot Morris :2024年民调是否成功克服了此前低估特朗普支持率的问题,需要结合投票数据进一步分析,以确定是否存在对特朗普支持者的非回应偏差。民调机构可能难以解决低估特朗普支持率的问题,这可能是特朗普特有的问题,而非普遍性问题。一些新兴的民调方法,例如“河流抽样”,比传统的在线面板和电话民调更准确。目前尚不清楚按以往投票情况进行加权是否有效,需要进一步研究。所有民调现在都是模型,没有黄金标准的民调方法。对民调结果的解读应谨慎,避免过度解读单一年的结果。

Deep Dive

Key Insights

Why did pollsters breathe sighs of relief after the 2024 election?

Polls were less error-prone in 2024 compared to 2016 and 2020, with state-level polling being the most accurate in at least 25 years.

What was the average polling miss in the 2024 election?

The average polling miss was 2.7 percentage points nationally and 2 percentage points in battleground states, significantly better than the 4 percentage point average over the past 25 years.

Why might pollsters have underestimated Trump's support in the 2024 election?

Pollsters might have faced non-response issues among Trump supporters, a problem that has persisted since 2016 and 2020.

Which polling method performed the best in 2024?

River sampling, which drives people to answer polls through social media ads, had the lowest bias and error.

Why did some pollsters choose to weight by past vote in 2024?

Weighting by past vote was seen as a way to ensure a proper mix of past Trump and Biden voters, though it's not clear if this method will be effective in future elections with different candidates.

What challenges do pollsters face with low response rates?

Low response rates force pollsters to model the electorate, making all polls essentially models rather than representative samples.

How did polls perform in measuring subgroup movements in the 2024 election?

Polls did a good job showing movement among Black and Latino voters towards Trump and among young voters, though some overestimated movement among Black voters and underestimated among Latinos.

Are live phone polls still considered the gold standard?

While some pollsters still value live phone polls, the rise of new methods like river sampling suggests that the gold standard may be shifting to more innovative approaches.

What should people take away from the 2024 polling results?

All polls are now models due to low response rates, and there is no longer a single gold standard method. Instead, there are good pollsters who use effective modeling techniques.

Are pollsters optimistic about the future of gauging public opinion?

Yes, pollsters are optimistic as polls continue to provide valuable insights into public opinion and democratic processes, despite the challenges posed by low response rates.

Chapters
The podcast discusses the performance of polls in the 2024 election, comparing it to previous years and highlighting the accuracy of state-level polling.
  • Polls were less error-prone in 2024 compared to 2016 and 2020.
  • State-level polling was the most accurate it’s been in at least 25 years.
  • Polls underestimated Trump's support for the third consecutive election.

Shownotes Transcript

Election Day was a moment of truth for pollsters. After high-profile misses in 2016 and 2020, and with a public that has become less trusting of polling, 2024 had the potential to be a make-or-break year for the polls.

Two weeks later, pollsters are, on the whole, breathing sighs of relief. Polls were less error-prone this year than in 2016 and 2020. By one measure, state-level polling was the most accurate it’s been in at least 25 years.

But that’s not the whole story. In this installment of the 538 Politics podcast, we evaluate how the polls did, covering the good, the bad and the statistically insignificant. Hold on to your priors.

Learn more about your ad choices. Visit podcastchoices.com/adchoices