A poll asks a sample of people, usually around a thousand, how they intend to vote, then scales the answers up to estimate the whole country. Good polls try hard to make the sample representative. Every poll has a margin of error, so small differences between polls often mean nothing at all.
A poll asks a sample of people, usually around a thousand, how they intend to vote, then scales the answers up to estimate the whole country. Good polls try hard to make the sample representative. Every poll has a margin of error, so small differences between polls often mean nothing at all.
A sample stands in for the whole population.
Representative means it looks like the country.
Margin of error is usually around three percent.
A poll is a snapshot of now, not a prediction of November.
If the sample is chosen properly, a small group really can reflect a big population, in the same way a spoonful of well-stirred soup tells you about the pot. The hard part is the stirring: reaching people who do not answer phones, young voters, and people in remote places.
A margin of error of three percent means a party polling at 32% could really be anywhere from about 29% to 35%. So a party 'rising two points' since last month may not have moved at all.
This is the single most misreported number in election coverage.
Different pollsters ask different questions, reach people in different ways, and adjust their raw numbers differently. Comparing one company's polls over time is more useful than comparing two companies on one day.
Have polls been wrong before?
Yes, in several countries, usually when the sample missed a group of voters.
Is our site poll scientific?
No. Ours is self-selecting, which means only people who chose to click took part.