Understanding Margin of Error
What the margin of error actually means, why polls within the margin should be called toss-ups, and how media routinely misinterpret this statistic.
What the Margin of Error Actually Means
A poll showing Candidate A at 48% and Candidate B at 46% with a +/-3% margin of error does NOT mean A is winning. It means A's true support is likely somewhere between 45% and 51%, and B's is likely between 43% and 49%. The ranges overlap, meaning B could actually be ahead. The race is a statistical tie.
The margin of error, technically the 'sampling error at the 95% confidence level,' tells you the range within which the true population value falls 95 times out of 100. It accounts only for sampling error, the randomness inherent in drawing a sample. It does NOT account for other sources of error: question wording bias, non-response bias, weighting errors, or late voter shifts. The real uncertainty in any poll is always larger than the stated margin of error.