Whose error?

What is “margin of error”? We see this term used in surveys and political polls quite often. What does it mean and how is it obtained? Whose error?

From a book often used in the Herald-Leader newsroom titled Numbers in the Newsroom:

Most organizations cannot survey an entire population, so they “sample” part of the population. Margin of error, also called sampling error, is a method for telling the reader how likely the poll results are correct. An example: A 2.5% margin of error means the the poll results will report answers within a 5 % point range. So, “error ” refers to the possibility of error within the survey. A detailed explanation of this involves statistics, formulas, confidence intervals, more. It’s important to know that margin of error is made up of three elements:

1. The number of people in the survey (the more people surveyed, the smaller the margin of error)
2. The differences in the answers (there’s a smaller margin of error when 90% respond one way than when 55% respond that way)
3. How accurate the pollster wants the answers to be

A simple way of understanding the concept: the smaller the margin of error, the more likely the results are correct.

The Denver Post has an interesting article on the current state of political polling.