Pollster ratings are grades that data analysts give polling firms based mostly on how accurate their past election polls were, with credit for being transparent about their methods. The best-known ratings today come from Silver Bulletin, which grades pollsters from A+ to F. A high grade means a firm’s polls have been closer to actual results than most, so it deserves more weight. It doesn’t mean its next poll will be right.
Ratings are useful for one quick job: deciding how seriously to take a poll you’ve never heard of.
Who rates pollsters now that FiveThirtyEight is gone? #
For years, FiveThirtyEight’s ratings were the standard reference. 538 switched from letter grades to a star scale in 2023, and then ABC News shut the site down in March 2025. Its ratings are no longer updated.
The main source now is Nate Silver’s Silver Bulletin pollster ratings, which grow out of the system Silver built at 538. Its January 2026 update ranked The Washington Post, Marquette University and NYT/Siena at the top. Because 2025 was an off-year with few polled races, that update changed little from the previous edition, which followed the 2024 election and gave AtlasIntel an A+ and the top overall rank for the first time.
How are pollster grades calculated? #
The details are statistical, but the ingredients are easy to follow.
1. Accuracy against real results #
The core measure compares each pollster’s final polls with certified results across many races and years. The error is adjusted for difficulty: polling a House primary with a low, unpredictable turnout is harder than polling a presidential race in a big state, and a pollster isn’t punished for taking on hard races. Silver Bulletin’s main score, Predictive Plus-Minus, estimates how accurate a firm is likely to be compared with an average pollster in the same races. A negative score is good.
2. Transparency #
Firms get credit for belonging to the American Association for Public Opinion Research (AAPOR) Transparency Initiative or sharing their data with the Roper Center archive. Pollsters who publish question wording, dates, sample sizes, weighting and crosstabs are easier to check, and ratings reward that.
3. Signs of herding #
Some pollsters, consciously or not, adjust their results toward the polling average so they don’t stand out. Silver Bulletin applies a penalty when a firm’s polls sit closer to the average than random sampling alone would allow. Our explainer on polling herding shows why that matters.
4. How much data there is #
A firm with a few dozen polls gets a rougher grade, shown as something like “A/B,” than a firm with hundreds. Pollsters found to have faked data or been banned get an F.
What about partisan lean? #
Ratings measure bias, a firm’s average lean toward one party compared with the actual results, separately from house effect, its lean compared with other polls in the same races during the campaign. A firm can have a noticeable house effect and still be accurate if it turns out to be right about who shows up. See what a pollster house effect is for the difference.
What do the grade tiers mean in practice? #
| Grade range | What it usually signals | How to treat a new poll |
|---|---|---|
| A range | Long record of low error, transparent methods | Take it seriously, even if it breaks from the pack |
| B range | Solid record, sometimes cheaper methods needing heavier weighting | Useful, especially as part of an average |
| C range | Larger past errors, a notable lean or less disclosure | Read alongside other polls, not on its own |
| D to F, or banned | Big misses, little disclosure or proven problems | Mostly ignore it |
| Unrated | Too few polls to judge | Look at its methods and sponsor instead |
Grades describe the pollster, not a particular poll. An A-rated firm sponsored by a campaign is still running an internal poll, and a C-rated firm can land close to the result.
Why isn’t a high grade a guarantee? #
Ratings look backward. Two recent examples show the limits.
The Selzer & Co. Iowa Poll had one of the best records in the country for decades. Its final 2024 poll showed Kamala Harris ahead in Iowa 47% to 44%. Donald Trump won the state by 13 points, a miss of about 16 points, far larger than any previous Selzer poll. Ann Selzer, who had planned to retire, announced her retirement from election polling later that month.
The polling environment also changes faster than ratings can. Phone response rates have fallen to low single digits, and methods that worked a decade ago need heavy adjustment now. A firm’s grade reflects how its methods performed under past conditions, not how they’ll hold up in the next cycle.
How should you use pollster ratings when reading polls? #
- Look up the pollster before reacting to a surprising poll. An unfamiliar name with no track record deserves caution.
- Check the sponsor. Is the poll paid for by a campaign or party group? That makes it an internal regardless of the firm’s grade.
- Read the method. Live phone, text-to-web, online panel or mixed? Likely voters or registered voters?
- Look at the average. One high-rated poll can still be an outlier, and an average of several decent polls usually beats any single poll.
For step 4, Election Tracker shows every published poll in a race with its pollster named on the card, next to the race’s 30-day average. The app doesn’t grade pollsters. Its average treats each pollster the same, counts a firm’s repeat polls in a race only once at its strongest sample, and weights newer polls more. So the pollster name on each card is your cue to check a rating when a poll looks out of line.
Frequently asked questions #
Can a low-rated pollster get an election right when a top-rated one misses? #
Yes. Any single poll can land close by chance or miss by chance. Ratings summarize long-run performance across many races, which is why they’re a better guide to the next poll than any one lucky or unlucky result.
Are online polls graded lower than phone polls? #
Not automatically. Grades are based on results, not method. Several online and mixed-mode pollsters score well because their results have been accurate, while some phone pollsters have slipped as response rates fell. Our guide to opt-in online polls covers where web polling works and where it struggles.
Why don’t top pollsters poll every race? #
High-quality polls cost a lot, so sponsors spend where the news value is highest, which usually means presidential swing states and marquee Senate races. Less competitive states and most House districts get few polls, often from less-established firms.
Where can I find a pollster’s grade? #
Silver Bulletin publishes its full ratings table on its site. For a quick check of a poll’s credibility without a grade, our guide to telling if a political poll is reliable lists what to look for.