Replies

Great POLL story in todays Suntimes

Few pluses, many minuses to early poll-watching

March 9, 2004

BY RICHARD ROEPER SUN-TIMES COLUMNIST Advertisement

Just three months ago, the media were all but ordaining former Vermont Gov. Howard Dean as the Democratic presidential nominee -- and we had the polls to back us up.

"Word that Howard Dean is picking up the endorsement of ... Al Gore came as Dean was already moving into a clear lead over his rivals," reported USA Today on Dec. 9, 2003.

" 'This is the first time in our national polling that we've seen Dean pull ahead,' said Frank Newport, editor in chief of the Gallup poll. He said some of the other candidates, notably Massachusetts Sen. John Kerry, show 'rapid sinkage.' "

Rapid sinkage? I think Mike Ditka can help with that.

But that December poll did indicate that Kerry was toast. He had 7 percent support from likely Democrat voters, putting him just two percentage points ahead of Carol Moseley Braun and four points ahead of Al Sharpton. And that's with a margin of error of +/- 5 percent.

Fast forward to March 2004. Kerry's the sure nominee and Dean's at home. As the saying goes, Dean was doing great in the race until it was time for people to actually start voting.

That's the thing about those polls: they're kind of a racket.

Of course, the pollsters will tell you they're not in the prediction game, they're in the windsock game. These surveys aren't designed to tell us who's going to win; they let us know which way the political breeze is blowing at any given moment. And because the media love to report on elections as if they're horse races, we breathlessly give you the results of the polls week after week. So from now until November, you'll be hearing story after story that goes something like this:

"In the latest poll results, Democratic contender Joe Candidate has surged ahead of Republican Senator Molly Incumbent.

"Thirty-eight percent of registered voters who didn't hang up on the pollsters said that if the election were held today and they weren't late for work and it wasn't raining, they'd vote for Joe Candidate, primarily because he was so funny on 'Conan' the other night.

"Thirty-four percent said they prefer Molly Incumbent, 20 percent are undecided and eight percent said they had never heard of either candidate but did know the names of all the remaining contestants on 'The Apprentice.'

"The poll has a margin of error of plus or minus five percent."

Filling in the margins

Let's examine that whole "margin of error" thing. Does "plus or minus five percent" mean Joe Candidate might have only 33 percent of the vote, while Molly Incumbent could have as much as 39 percent? Or does it mean the 4 percent differential, when coupled with the five percent margin of error, means the race is the dreaded "statistical tie"?

I love that term, "statistical tie." As opposed to what, an artistic deadlock?

Beyond that, doesn't the phrase "margin of error" imply there's a way of checking the poll against a survey of, like, everybody?