Polling

Traditionally, the margin of error is used with 98% of probability.
That is, the probability of the sample erring by more than "z" percentage points is 2%. 98% of the samples would fall inside the "z" percentage gap.

fbaezer-
Thank you. But how does the pollster know that his poll reaches that standard?

Through laws of statistics.
For example a 100 person sample has a +/-10% error range with 98% of proobability. The error range is only about 2% in a 1500 person sample. Oddly, the size of the polling universe matters little. Doesn't vary much if your polling a small county or a big country.

Now, the assertion is correct IF the sample is truly aleatory. That is, if everyone in the universe had the same chance of being included in the sample.
Usually there are some "house bias" and the true margin is a bit higher that the stated one.