Read Polls With Care

With polling data as common in the news pages nowadays as box scores are on the sports pages, thoughtful readers need to approach these compiled numbers with caution.

"The first question to ask is: 'Are the questions quoted exactly as they were put to the respondents?' " says Renate Kcher, managing director of the Allensbach Institute here.

In polling, the formulation of the questions is all-important, Dr. Kcher says. If a news article paraphrases poll questions, critical nuances can be lost. While agreeing that it can be tedious to reproduce a questionnaire verbatim, she says footnoting to the original source is a help.

"You should also pay attention to whether 'undecided' responses are included," Kcher says. "The more complicated the question, the more 'undecideds' there should be. And if these are eliminated, and results are presented in terms of for or against, you get a distorted picture."

Some researchers see the size of the sample - a minimum of 1,500 people, for instance - as of great importance. But for Kcher, it is less vital.

Responses to the so-called "Sunday question" - "If the election were this Sunday, whom would you vote for?" - are a staple of the weekly press in Germany. For some time they have suggested that the government of Chancellor Helmut Kohl is in trouble with voters.

But these data are often misunderstood, Kcher says. "Chancellor Kohl says, 'Others win the polls, but I win the elections,' " she notes.

"Popularity figures for politicians are always suspect. The leader always polarizes, drawing approval from some but disapproval from others.... Someone who's not in power has more opportunity to make a purely positive impression without the negatives."

With the next federal elections nine months off, it is simply too early for "Sunday" data to be relevant.

"You can't project a year ahead," Kcher says. "At this point only about 40 percent of the electorate has decided whom to vote for."