2024 has fewer polls, but they are higher quality
As Election Day approaches, many of you may find yourselves obsessively checking the polls, poring over every minute change in our forecasts and averages to try to glean any new insights you can about the state of the race. We get it — looking at the polls is kind of our thing here at 538. But all that obsession over tiny shifts can get pretty exhausting, so today we decided to leave the horse race behind (just for a moment!) and zoom out to look at the bigger picture of the polling landscape this year.
Each election cycle brings its own challenges for the polling industry, from the steady transition from live phone to online polling, to plunging response rates that make it harder than ever to get a representative sample. All of these challenges (combined with less-than-stellar performances in 2016 and 2020) can make it easy to view political polling as stuck in a downward spiral.
But our analysis shows that there are more reasons to be optimistic than pessimistic about the state of the polls in 2024. While we've seen a decline in the volume of polls, there are still plenty of encouraging signs that make us confident in the data we have.
Fewer polls and fewer pollsters
Let's start with some of the bad news: We have fewer polls this year than in previous presidential cycles. While we've collected more Senate polling than in most prior elections, the number of presidential and House polls has dropped notably from past cycles.
We've also seen a decline in the number of unique pollsters who have released at least one poll. So far this year, we've collected 1,532 national, state and district-level polls from 209 pollsters conducted between early May and late October; 1,301 of those polls included a presidential question, while 683 of them asked about at least one congressional race. (For more information about how we analyzed this data, see the methodology statement at the end of this article. And if you want to dive in yourself, you can download all the data we used for this story.)
This isn't the first cycle we've seen declines in the number of polls and pollsters: The 2022 midterms saw the fewest polls in a midterm since at least 2010, and the polls we did have were the most partisan on record. On the surface, it looks like 2024 may be a repeat of two years ago. But a closer analysis reveals that there's much more to the story than the topline numbers — and a lot fewer reasons to panic.
The polls we do have are (probably) better
One possible explanation for this drop in the total number of polls is that the most prolific pollsters of 2024* are actually a bit less prolific than those of the past. There are only 10 pollsters that have released more than 25 surveys this cycle, fewer than 2020 (13 pollsters) and 2016 (11 pollsters). Moreover, these 10 most prolific pollsters account for a much smaller percentage of the data than they have in the past. In fact, this decrease in polls from the 10 most prolific pollsters accounts for a significant share (if not all) of the difference between the total number of 2024 polls and previous cycles; if the 10 pollsters that published the most surveys are removed from the dataset in each cycle, 2024 has more polls than two of the previous three cycles, and is behind 2020 by just 17 polls.
This is probably good news: Having a small number of pollsters contributing a large share of the surveys likely doesn't help give a broad picture of the race. In 2016, for example, the 10 most prolific pollsters accounted for 58 percent of all polls released during that cycle, which could cause more careless aggregators to put too much weight on the same methodology.
In addition, the number of polls in a cycle may be less important than the quality of those polls. Despite the gloomy 2022 pre-election polling outlook, polls ended up performing exceptionally well that year. And this year, there are signs that the polls may be of higher quality than in past presidential cycles.
For example, things look a lot better when we consider the 538 pollster rating of each poll this year, which can provide useful context about how much we can trust the data from a particular pollster. Fewer polls come from unrated firms than in the past, and the average pollster rating of polls for which we have historical data is also higher.
Part of the story here is that 2024 has seen a drop in the number of new pollsters, defined as pollsters for whom we hadn’t seen horse-race polling in previous presidential cycles and for whom we had no pollster rating based on past work. In 2024, we have 57 firms in our database that meet that definition, compared to 65 or 66 in each of the previous three cycles.
And while it may feel to some observers that we’ve seen an influx of new pollsters flooding the averages with data, the new pollsters of 2024 are fairly similar to those of previous cycles: On average, they’ve published 4.9 surveys, a bit less than the 5.4 surveys such pollsters put out in 2020 and a bit more than the 4 surveys new pollsters published in 2016.
Another sign that the polls may be higher quality this year: The 2024 congressional polls are less partisan than in past cycles (more on the presidential polls in a bit). This year is on track to have the lowest share of partisan House polls since at least 2008, at 61 percent. A similar story is playing out in Senate races, where only 16 percent of polls released this cycle have come from partisan sources. As a result, around a quarter of all Senate and House polls have a partisan flag, the lowest mark in the last four presidential cycles.
On the Senate side, this decline represents a more than 50 percent drop from 2020, when 35 percent of polls came from partisan sources. Still, the 2024 Senate mark is closer to what we saw in 2012 and 2016, when 14 and 20 percent of Senate polls, respectively, were partisan.
And while having a majority of House polls hail from partisan groups might seem like a lot, that's an unavoidable reality with this kind of polling. After all, most of the organizations who are actually interested enough (and well-funded enough) to poll individual district-level races are usually campaigns or party-affiliated groups. While it's still a high number, 61 percent is much lower than, say, in 2016, when about three out of every four (!) House polls came from a partisan source.
A smaller share of partisan House and Senate polling means that their results are less likely to be skewed toward one party or the other. Of course, our models already account for potential partisan bias, but the less you have to adjust a poll, the less room there is for error to creep in. That's good news for the overall accuracy and reliability of congressional polling averages and forecasts, and particularly beneficial for downballot races where we have less data to work with.
Beyond partisanship, the kinds of organizations that are sponsoring and conducting polls matters, too. So far this cycle, 50 percent of polls have been sponsored or conducted by media groups or universities. That's in line with the 55 percent from 2020, but much higher than prior cycles; in 2016, only 38 percent of polls came from media or university sources, and in 2012, it was just 30 percent.
Media organizations and universities are usually more rigorous, transparent and nonpartisan than other kinds of pollsters and sponsors. Just take a look at our pollster ratings — eight of the top 10 ranked pollsters are news organizations or universities. The fact that these groups are making up a large share of the polling this year is a good sign.
When it comes to the presidential election, we also have some good news: There is a lot of swing-state polling this time around, despite some assertions otherwise.** While the total number of polls in swing states has remained relatively stable, even increasing a bit from some previous cycles, the fact that so few states appear to be competitive this year has meant that pollsters have been able to focus their resources, yielding far more polls in each swing state than past cycles.
In addition to having, on average, more polls per swing state, none of the competitive states have been overlooked, as we've sometimes seen in the past. The swing state with the fewest polls this cycle is Nevada, where 69 polls have been conducted. In 2020, on the other hand, two states rated as lean Democratic, Nevada and New Hampshire, both had fewer than 20 surveys conducted in the run-up to the final weeks of the campaign. And while 69 polls in Nevada may be the fewest this year, that's more than any competitive state got in 2012: the most-polled state in that cycle was Florida, which topped out at 66 surveys.
Not everything is coming up roses, though
On the other hand, there are some troubling trends we first observed in 2022 that have continued this cycle. While the number of polls for House races has been steadily dropping for years, we're also seeing a drop in the number of competitive races*** that pollsters have been focusing on. Of the 43 competitive races in 2024, 74 percent have been surveyed more than once, lower than both 2016 and 2012 and continuing a dropoff we first saw in 2020, but also observed during the 2022 midterms. And the total number of polls in competitive districts is the lowest we've seen in the last four presidential cycles.
One possible reason for this drop in polls of competitive races is a significant drop in partisan polls of these seats. There have been only 81 partisan surveys in competitive House seats, compared with 129 in 2020, 111 in 2016 and 157 in 2012. So while this makes the 2024 polls the least partisan overall (just 64 percent of 2024 House polls in competitive seats have a partisan sponsor), it may mean that internal campaign polling is being withheld more often than in past cycles. This gives us less insight into how the campaigns are running their races, as well as less data on not just the horse race, but issues impacting each district.
And while congressional polling has been noticeably less partisan this cycle, the decline in partisan presidential polling has not been as steep. This cycle, 14 percent of presidential polling has come from a partisan source, a slight decline from 2020, but still noticeably higher than in 2016 or 2012.
And while Democrats released the majority of partisan presidential polls from 2012-2020, the reverse is true this year, with Republican sources releasing more partisan surveys — though the split is still relatively close. Rather than a huge influx of new Republican pollsters and sponsors, the change seems to be driven mostly by Democratic organizations publishing fewer surveys; 11 percent of 2020's presidential polls had a Democratic sponsor, compared with just 5 percent of presidential polls in 2024.
We don't currently expect this to have a huge impact on our forecast and averages, however. As mentioned above, we account for a poll's partisanship when incorporating it into our models. The overall share of partisan presidential polls remains quite low, and the gap between the share of partisan polls released by each party isn't significantly different than in 2020. Given all of this, we aren't all that worried about these surveys tipping the scales too much in either direction.
* * *
The bottom line is, everything's OK. Even though we have a bit fewer polls than some recent cycles, the ones we do have are probably a little higher quality than in the past. We're seeing more data from sources we already know (and have a sense, based on our pollster ratings, of how much to trust) and fewer individual pollsters dominating the aggregates. Also, the cycle isn't over: We have another week to add to this data, so when all is said and done, we may end up exceeding some previous cycles (we're only 108 polls away from matching 2020's output when both cycles had 15 days to go, and in the last week of the campaign, we'll see a lot of new data). So the next time you have the urge to panic about the state of polling in 2024, take a deep breath — things aren't as bad as they may seem.
Footnotes
*The 10 most prolific pollsters of 2024, in order of the number of surveys published, are Redfield & Wilton Strategies, Morning Consult, ActiVote, Emerson College, YouGov, The New York Times/Siena College, RMG Research, Public Policy Polling, Cygnal and InsiderAdvantage.
**We define a swing state as any state rated as Lean Democratic, Lean Republican or Toss-up by the Cook Political Report as of Oct. 28 at 9 a.m. Eastern. For previous election cycles, we include any state that had one of those ratings as of the day before the election.
***We define a house race to be competitive if it is rated as Lean Democratic, Lean Republican or Toss-up by the Cook Political Report. For previous election cycles, we include any race that had one of those ratings as of the day before the election.
Methodology
This analysis is based on all non-tracking polls with a final field date between 180 and 15 days before the presidential election in each of the last five election cycles, as of Oct. 28 at 9 a.m. Eastern. Polls are included if they ask at least one question about who respondents will vote for in either a House, Senate, presidential or gubernatorial general election or "jungle" primary, or if they ask at least one generic ballot question.
Polls are classified as partisan if they were conducted by or on behalf of a candidate, party, campaign committee, PAC, super PAC, hybrid PAC, union, corporation, 501(c)(3), 501(c)(4), 501(c)(5) or 501(c)(6) organization that conducts a large majority of its political activity on behalf of one political party. Polls are classified as internal if they were conducted on behalf of a candidate, party, campaign committee or other official party apparatus.
Pollster ratings for each cycle are calculated according to our standard pollster ratings methodology, but include only polls that were conducted prior to the relevant cycle.