ActiVote’s 2024 Polling Recap

Many people report that they have never participated in a poll, while they would love to have their voices heard. With ActiVote, that has changed: anyone has the chance to participate! When our users prepare for their election in the ActiVote app, they also know that their opinions may be (anonymously) included in our polls.

In 2024, ActiVote published a total of 235 polls, with the vast majority in the lead up to the presidential election on November 5:

According to Mary Radcliffe et al. from 538, we were the 3rd most prolific pollster of this cycle.

Comparing the accuracy of pollsters is a complicated task which ultimately is left to 538, who will likely publish their updated pollster ratings sometime in Q1, 2025. In the meantime, we can look at it from a few angles to get a first impression.

The National Popular Vote

While a few votes are still being counted, it looks like Trump will win the popular vote by approximately 1.4%.

Our last poll, published November 2nd, had Trump +1.0%, thus a minimal error of just 0.4%, which makes us the 2nd most accurate pollster in terms of getting the national popular vote right. In comparison, we can see how “the field of all pollsters” did, by looking at the polling aggregators. 538 had Harris +1.2%, with an error of 2.6%, while RealClearPolitics had Harris +0.1%, thus a 1.5% error.

Some may think that we just got lucky with our last poll, but if we look at our national polls throughout the election cycle, we have routinely shown the trends ahead of others.

Biden dropped out on July 21, 2024 and by July 23 we showed Trump leading by 1%. Two days later NYT/Siena College published the same result. What followed was a two-month period in which Harris took the lead in our polls and in the polling averages. However, by early October it became apparent in our polls that Trump was retaking the lead. Our October 9 poll showed Trump +1.2%, which was the first National Poll since September 22nd from any pollster showing a Trump lead, at a time that the 538 average still showed Harris +2.5%. In other words, ActiVote was a leader in showing a significant change in the race.

Swing States

Now let’s look at the all-important swing states, of which we polled all but Nevada. The table below shows the actual result per state, followed by the ActiVote poll and error for that state, and then the 538 average and error for that state. Red numbers indicate a Trump lead, while blue numbers indicate a Harris lead.

ActiVote did exactly as well as the 538 averages, with an overall error of 2.1%. ActiVote correctly called 4 out of 6 states for Trump, while the 538 averages called 3 out of 6 correctly. In other words, ActiVote as an individual pollster did as well or better than the combined wisdom of all pollsters as aggregated in a polling average.

Performance by Top Rated Pollsters

The 538 pollster rating awards high ratings to those pollsters who produce accurate polls while being transparent about their methods. We wanted to see how we compared to the best rated pollsters in this election cycle.

To this end we restricted our sample of pollsters to those who produced at least 10 polls in this cycle that will be included in this years ratings (to ensure a statistically relevant sample) and have a transparency score of at least 8 out of 10 (ActiVote’s score is at least 8 of 10 as well, according to 538’s article about ActiVote).

The table shows that Emerson (rated 10th before this cycle), leads the charge, closely followed by ActiVote.  It is noteworthy that 4 of the top-10 pollsters (Siena/NYT, YouGov, Marist and UMass Lowell) had significantly larger errors this cycle.

Conclusion

ActiVote has been successful at ensuring that voters have a place to express their views on candidates and issues. Each of the various ways of measuring polling accuracy shown here indicate that in this year of prolific polling, ActiVote easily held its own in comparison with the best established and highly rated polling organizations.