Margin of Error
This article includes tips on how to read polling data, within the context of how margin of error and polling samples work.

Chance Error and Bill Clinton's Political Fortunes
This article, written during the 1996 election, describes how margin of error may explain polling data disparity.

Statistics Every Writer Should Know: Margin of Error
This page, at the "Statistics Every Writer Should Know" site, points out the importance of paying attention to margin of error statements in political polling.

What Is a Margin of Error?

Margin of error is one of the most difficult statistical concepts to understand. Simply stated, a margin of error states pollsters' confidence in the data they have collected. Say that during the Clinton/Dole race, we heard the following news report: A Gallup poll conducted yesterday indicates that, if the election were held today, 70% of voters would choose Clinton, with a margin of error of +/- 5%. What would we know from this statement?

What a Margin of Error Tells Us
The report actually says that 70% of those polled said they would choose Clinton. If the pollsters were to poll another group of people (different people than the ones in the first sample, but with the same demographics as the first sample), also at random, they could bet, with a significant amount of certainty, that no fewer than 65% (the minus 5% in the margin of error) and no more than 75% (the plus 5% in the margin of error) of the new group would say they would choose Clinton.

How Do They Know That?
Now, you might ask, "How do they know that's what would happen?" In fact, margin of error is based solely on a mathematical formula. Pollsters have been checking public opinion since the beginning of the century. Over time, mathematicians began to see patterns that would reoccur each time polling was done. Mathematical formulas designed to analyze statistics are the result of those observations.

Remember the blood example. The more blood cells you collect, the better the chance will be that you'll have a truly representative sample of the entire bloodstream. The same thing happens with polling. The more people a pollster talks to, the smaller the margin of error will be. Mathematicians, with hundreds of years of experience studying chance, have figured out a formula that tells the pollster just how many people he or she has to poll to to get a particular margin of error.

Think About It
Think about this. Remember the last Journal-Times report that said "Higgins is catching up with Fletcher, though Fletcher continues to have a slight lead of 3%, with a margin of error of +/- 5%"? Do the statistics really say that Fletcher has a slight lead? Think about it! What do they really say?

Three Months Before the Election