Follow the data
Polls & political data
We don’t publish our own horse-race numbers — we point you to the most credible, transparent sources and help you read them well. Here’s where to follow the polls, and how to avoid getting fooled by them.
Poll averages & forecasts
Start with averages, not single polls. Each of these publishes its methodology.
- FiveThirtyEight (ABC News)Poll averages and forecasts with published methodology and pollster ratings.Nonpartisan
- The New York Times / Siena pollsHigh-quality live polling and averages of key races.Center
- RealClearPolitics averagesLong-running averages that aggregate a broad range of public polls.Center
- Silver Bulletin (Nate Silver)Model-based election forecasting from FiveThirtyEight's founder.Nonpartisan
- 270toWinInteractive electoral maps, historical results, and consensus race ratings.Nonpartisan
- Split TicketData-driven race ratings and election analysis.Nonpartisan
Nonpartisan race ratings
For House, Senate, and governor races, ratings (“Toss-up,” “Lean D,” etc.) often tell you more than raw polls.
Public opinion & research
Beyond the horse race: what Americans actually think about issues, and how it’s changing.
- Pew Research CenterDeep, transparent public-opinion research on politics and policy.Research
- GallupLong-run trends on approval, party identification, and issues.Research
- AP-NORCRigorous surveys from the Associated Press and NORC at the University of Chicago.Research
- PRRINonpartisan research on religion, culture, and public policy.Research
How to read a poll without getting fooled
One poll is a snapshot, not the story
Any single poll can be an outlier. Trust averages of many polls over any one headline number — especially one that seems surprising.
Respect the margin of error
A 3-point lead with a ±4-point margin is effectively a tie. Small movements within the margin usually aren't real change.
Check who was surveyed
Likely voters, registered voters, and all adults can give very different results. Late in a race, likely-voter polls are usually more predictive.
Watch the timing and the trend
A poll taken right after a big news event can bounce. A steady trend across weeks means more than a single fresh number.
Know the pollster
Some pollsters have strong track records; some have a consistent 'house effect' that leans one way. Ratings from FiveThirtyEight help you weigh them.
Polls measure now, not November
A poll is not a prediction. It captures a moment — turnout, undecideds, and events still decide elections.