Sometimes in Polling, It's All in the Question

WSJ.com - The Numbers Guy: "as noted on Mystery Pollster, are surveys about impeachment – the site's Mr. Blumenthal points out that many members of the public aren't sure of the definition of the word "impeachment" -- and one about prospective Iraq policy.

By contrast, there are some issues for which the wording of polling questions may not have much impact. Dr. Jones cites a poll last July in which Gallup split up participants into two groups. One was given an explanation of Roe v. Wade and asked if they would like to see the Supreme Court "completely overturn" the decision. The other group was asked, simply, "Would you like to see the Supreme Court overturn its 1973 Roe versus Wade decision concerning abortion, or not?" The results in both groups were virtually identical -- a two-to-one margin against overturning the decision.

Part of the problem is that survey respondents are eager to seem helpful, even if it means claiming an opinion they don't hold firmly. Dr. Jones noted that some experimental surveys have asked respondents about a nonexistent policy with a fictional name, and found that many people offered opinions. "They want to give you what you're asking for, which is an opinion, even if they haven't thought about it," Dr. Jones says."