Friday, June 19, 2009

Tips on Arguing: Margin of Error

Inquiry:

Ever notice the way that news organizations will pelt you with polling data during an election cycle, and then wonder why the actual results defy the polls? There are a lot of reasons this happens, but one of the most pervasive is due to a misunderstanding about something called the margin of error.

Every survey contains an inherent margin of error that is calculated using methods developed by statisticians. The primary reason for doing this is that survey-takers only talk to a portion of the total population, and they can never be completely sure that those people are representative of the whole group. Some people lie. Some people forget to vote. Others change their minds. And there’s always the possibility that the people you didn’t ask would have given you completely different answers.

Figuring out what the margin of error is in a given case requires some mathematical background. Fortunately, this is often done for us ahead of time, and knowing how to interpret the result is a simple process that takes less than a minute. It’s so easy to do that any reporter who fails to account for the margin of error is practicing shoddy journalism.

So, let’s say that Bob and Carl are running for mayor of Blandeville. The night before the election, a poll of 500 registered voters reveals that 44% are in favor of Bob, and 51% are in favor of Carl (the rest are undecided), with a margin of error of ± (plus or minus) 4 percentage points. Carl must be a shoe-in, right?

Not necessarily. The margin of error shows that any of these numbers is likely to be four percent below or above what the pollsters determined. So, what the poll really says is that Carl’s chances may be as low as 47%, or as high as 55%. Likewise, Bob’s support may be anywhere from 40% to 48%.

Since Carl’s lowest score (47) is less than Bob’s highest (48), Bob may actually be ahead. Nobody who pays attention to this will be surprised if Bob ekes out a victory tomorrow.Don’t be fooled by people who ignore a margin of error. It can make all the difference.

No comments: