Margin of error

Margin of error expresses how much a sample-based estimate, most commonly a poll result, might differ from the true value in the full population purely because of sampling. A result reported as "52%, plus or minus 3 points" means the true population value is estimated to fall within that range, at a stated confidence level.

It is calculated from the standard error of the estimate multiplied by a factor tied to the desired confidence level, commonly about 1.96 for a 95% confidence interval, which is also how the margin connects directly to a confidence interval. Margin of error shrinks as sample size grows, but slowly: because standard error scales with the square root of sample size, cutting the margin of error in half requires roughly quadrupling the sample.

Margin of error matters because it is the standard way pollsters and researchers communicate sampling uncertainty, but it only accounts for random sampling variation. It says nothing about sampling bias, non-response, or a poorly weighted sample, all of which can push an estimate off in a consistent direction that no margin of error will reveal. The common misreading is treating a reported margin as the total possible error in a survey, when systematic bias is often the larger, unmeasured risk.