CupTalks Blog

A poll shows Candidate A ahead 48% to 45%, with a margin of error of about four points. Is the lead real? Not necessarily. That margin applies to each candidate’s share, not to the gap between them, which is roughly twice as uncertain. Larger samples and poll averages shrink random noise, but they can’t fix bias or errors that polls share. Add in who chooses to respond, how voters are weighted, and who actually turns out, and one number can hide a lot of uncertainty. Here’s how to read a poll, what each kind of uncertainty means, and the questions worth asking before you trust a headline.