The vagaries of political polling were on display throughout the primaries, but perhaps the most jarring example came last month when the Gallup Organization issued contradictory polls on the same day. Gallup's national poll conducted for USA Today showed Hillary Clinton leading Barack Obama by 7 percentage points. But at the same moment, Gallup's national tracking poll showed Obama leading by 5 percentage points.
How is that possible? The most cynical view would be that political polls are just cons and Gallup's gaffe simply confirms it. But professional pollsters point out that their industry's frequently misunderstood fine print, known as Margin of Error, provides a ready explanation when things go wrong.
Although Gallup's two polls differed by a total of 12 percentage points, they were both within the Margin of Error  meaning they were both statistically tied. Although the polls prompted diametrically opposite headlines, in Gallup's view they were each perfectly correct.
What exactly is Margin of Error? The question is addressed by prognosticator John Zogby, whose Web site explains: "Working with a panel of psychologists, sociologists, computer experts, linguists, political scientists, economists, and mathematicians, we explore every nuance in language and test new methods in public opinion research." Sure, but what about the actual Margin of Error? "If we do the same poll 100 times," Mr. Zogby explains, "in 95 cases out of 100 we will get the same results, plus or minus a certain percentage."
The key words therefore are "plus," "minus," and "a certain percentage." By taking all of these into account, deciders like Mr. Zogby determine what they call a "margin" by which they can avoid ever admitting an "error."
The beauty of what Mr. Zogby is talking about is lost on most pundits. Let's say candidate "M" gets 51 percent of the vote and candidate "O" gets 45 percent in a poll with a Margin of Error of plus or minus 4. Most pundits will confidently declare that M's lead is "outside the Margin of Error." But it's not. The Margin of Error applies to both numbers  so M's vote may be 47, and O's vote 49, in which case O is actually leading.
It gets worse. When Zogby mentions "95 cases out of 100," he means that 95 percent of the time their polling results will fall within the Margin of Error. But the other 5 percent of the time they won't, in which case the results may not even be as accurate  or inaccurate, as the case may be  as the other 95 percent of the time.
This should explain why: (a) the public has lost confidence in polls; (b) pundits are fearless in criticizing polling data, and (c) the winner on "American Idol" is rarely decided within a Margin of Error of plus or minus 100 million votes.
Still, we should all get behind the Margin of Error way of looking at life. I'd be perfectly willing to pay my credit card bill with a Margin of Error of minus $100. I'm certain I could remember my wedding anniversary within plus or minus 30 days. And imagine how successful the Giants and A's would be if they played with a small Margin of Error of, say, plus or minus four runs.
© Peter Funt. Originally published in The Monterey County Herald.
