<< back to the archive list

September 3rd, 2008

Obama and Osama / more on margin of error

by Barbara Wallraff

A few months ago, some people who follow politics got fussed because Microsoft Word’s spell-checker advised those who typed “Obama” to consider revising it to “Osama.” This was a conversation starter, or stopper, for sure. Still, it was obvious -- at least to me -- this was happening because Osama bin Laden had been major news for years before Barack Obama burst onto the scene. No malice, no conspiracy, no big deal.

So imagine my surprise the other day when I decided to review a transcript of Obama’s nomination-acceptance speech in Word and saw the spell-checker’s telltale squiggly red line under “Osama.” (In the speech, Obama said, “We must take out Osama bin Laden and his lieutenants if we have them in our sights.”) I clicked to see Word’s suggestion -- and now I was advised to consider revising “Osama” to “Obama”!

The switcheroo must have taken place during one of the periodic updates my computer installs automatically. I’m sure no malice, no conspiracy, was involved this time either -- but neither was much common sense on the part of the programmers. FYI to the Microsoft team responsible for stuff like this: There isn’t just one guy who keeps changing the spelling of his name. Barack Obama and Osama bin Laden are different people! Please make a note of it!

Mike Shpiece, of Farmington Hills, Mich., writes: “Your recent discussion of a survey’s ‘margin of error’ completely missed the meaning of that term.

“A survey asks questions of a number of people and projects those results onto the entire population. For example, a survey might ask 1,000 people who they plan to vote for in the next election. If 49 percent say Obama and 43 percent say McCain, the pollster would predict that the results of the election will be the same -- 49 percent to 43 percent.

“Can the pollster guarantee that the election will produce that result? No, for a variety of reasons. A lot can happen between now and November that could change people’s minds. Further, problems with the way the survey was conducted could affect the results. If, for example, the sample was drawn only from readers of The Wall Street Journal, we would expect the results to be different from a sample of voters in general. The way questions are asked might also change the results.

“‘Margin of error’ refers to something else. Even if everything was done perfectly and you called 1,000 people who will vote in November and the 1,000 people were chosen completely at random, it is possible you might call 1,000 registered Democrats. Obviously, that would bias the results. Statistically, the likelihood that all 1,000 calls were to Democrats is infinitesimally small, but as long as there are 1,000 Democrats somewhere in the country, it is possible. The margin of error says that statistically, there is a 95 percent chance that the ‘true’ number is at least that close to the reported number.

“So if the survey in our example had a margin of error of plus or minus 3 percent (which is about right for a sample of 1,000), there is a 95 percent chance that Obama would have between 46 percent and 52 percent support.”

Dear Mike (and the other readers who wrote to make the same point): Thanks for the lesson. That’s one mistake I won’t make again.

© Copyright 2003 by Barbara Wallraff. Reprints require prior permission. All rights reserved.

<< back to the archive list