What is the margin of error? We hear it all the time, especially during the political season when commentators talk about polls, and how far someone is ahead in the polls with a “plus or minus 3 percent margin of error.” Or maybe you’ve heard the Car Talk guys talk about their Director of Statistical Research, Marge Inovera.
Statistically speaking, it’s sometimes a bit easier to say what margin of error isn’t. Margin of error does not refer to the results of studies based on the quality of research, the sloppiness or excellence it is done with. That’s more along the lines of best practices and standards. Margin of error, mathematically speaking, is the range around any numerical answer that’s given in survey report and how closely the results represent the population on the whole. In defining it this way, you have to assume well-executed research methods using best practices.
Let’s try an example: We’re on the hunt for a new flavor of ice cream to make, and we wanted to know what base it should be made with, chocolate or vanilla. If we want to appeal to the greatest number of people, we would want to find out how many people prefer vanilla ice cream to chocolate.
Out of a population of one million people, we would need to survey 384 people to give us a 95% confidence level and a margin of error of plus or minus 5 points.
Stated another way, we are 95% confident that the answer we get from our sample is plus or minus 5 points of the answer we would get if we surveyed the entire population.
So if our survey results show that 65% of the people we polled prefer vanilla, we can say that between 60 and 70% of the general population will also like vanilla. (That’s because our margin of error is plus or minus 5 points: 65 minus 5 points is 60, and 65 plus 5 points is 70.)
And for our example, since more people prefer vanilla to chocolate — 60 – 70% versus 30 – 40% — our new flavor should be vanilla.