In the states that determined the election, polls generally showed losses of between 1 and 3 points. RCP averaged Trump in national popular vote ahead by 0.1 and will probably win by 1 or 2 points. An error of 1 to 3 percentage points for polls, which are blunt instruments that typically use fewer than a thousand interviews, to estimate how an entire state or nation feels. Great.

Polls in competitive Senate races were only slightly less accurate. Some results are yet to be finalized, but so far there have been only two troubling omissions in key Senate races: inflated predictions Democratic Senator Jacky Rosen of Nevada And Snubbing Texas Republican Senator Ted Cruz about 5 points each. 538 and other elections where the RCP calculates the average polls were missed by only a few points. Again, it’s unrealistic to expect polls to nail every outcome; An error of 1 to 3 points is as good as it gets.

• In the Nevada Senate race, the 538 average put Rosen ahead by 5.7 points over Republican Sam Brown, while the RCP average put him ahead by 4.9 points. Showing latest results Predicted winner Rosen leads by 1.4 points.
• In the Michigan Senate race, the 538 average put Democrat Elissa Slotkin ahead by 3.6 points over Republican Mike Rogers, while the RCP average put her ahead by 2.3 points. The results show Slotkin as the predicted winner. increased by 0.3.
• In Ohio, the 538 average put Republican Bernie Moreno ahead by 0.8 points over Democrat Sherrod Brown, while the RCP average put Republicans ahead by 1.7 points. The results show Predicted winner Moreno 4 points ahead.

• In Wisconsin, the 538 average put Democrat Tammy Baldwin ahead of Republican Eric Hovde by 2.2 points, and the RCP average put her ahead by 1.8 points. The results show Predicted winner Baldwin leads by 0.9 points.
• In Montana, the 538 average put Republican Tim Sheehy ahead by 6.9 over Democrat Jon Tester, and the RCP average put the Republican ahead by 7.7. The results show Predicted winner Sheehy earned 7.4 points.
• And in Texas, the 538 average put Cruz ahead by 4 points over Democrat Colin Allred, and the RCP average put Cruz ahead by 4.4 points. The results show Cruz, who was expected to win, gained 8.6 points.

These results are solid and will keep the survey industry alive. However, we do not have a clean bill of health yet.

Polls still plagued by non-response: Almost 99% A proportion of people selected for the survey did not complete the survey. Some of the groups we need to know the most about—young voters, Latino voters, the politically unaffiliated—are the hardest to survey.

Some pollsters can also be “guts.” This happens when an unprincipled pollster gets an unexpected result; he can throw it away or tinker with his statistical models until his polls match the average. This might explain why a suspicious number of polls in Pennsylvania show Trump and Harris exactly tied.

And while the polls weren’t wrong by much, they consistently undershot Trump by a few points. Ideally, surveys would be unbiased; Instead of just missing Trump voters, they were underestimating Harris by nearly half.

Pollsters have partial solutions to each of these problems. We use statistical tools such as weights to ensure that less responsive groups are given the right amount of influence. The same weights can create the illusion of a “herd” — meaning some pollsters adjusted their samples to get the right number of Trump and Biden 2020 supporters. naturally brought his results closer to the average. And although the polls missed some of Trump’s advantage in 2024, the error was significantly smaller than in 2020.

We do not have complete solutions to these problems. No one knows how to turn America into a nation of eager survey respondents, or how to prevent anxious, underfunded firms from unconsciously pushing their results toward the mean. And we couldn’t find the last tranche of Trump voting. But I hope that on Tuesday we gave our industry a little more time to solve these problems. I’d like to think we deserve it.