[lbo-talk] polls & wording

Doug Henwood dhenwood at panix.com
Wed Mar 3 11:21:27 PST 2004


THE POLLSTERS Mark Mellman

It's a question of wording

This will only look like a column about gay marriage, the Arctic National Wildlife Refuge (ANWR) and free speech. It's not. Rather it's an examination of something much drier and less highly charged: question wording.

Everybody knows that different questions often beget different answers. This is obvious when one looks at big differences. Last year when all the polls were showing opposition to oil drilling in the ANWR, an oil industry front group released a poll purporting to show that voters favored drilling in ANWR.

The actual question text did not ask about ANWR at all, however. It asked whether oil drilling should be allowed "in Alaska." Since no one I am aware of favors banning drilling in all of Alaska, the question, as worded, was irrelevant to the real public policy debate and likely designed to elicit the answer oil companies wanted. Indeed, that kind of deception was probably the only way they could show public "support" for their position.

But pollsters have long known that even very subtle differences in question wording can have a real impact on the answers. One of the oldest findings of this kind involves the difference between "forbidding" and "not allowing."

In the 1940's 75 percent said the U.S. should not "allow public speeches against democracy." Only 25 pecent said such speeches should be allowed. It follows then, that Americans wanted to forbid speeches against democracy, right? Not so fast.

Here people were much more divided. Twenty points fewer (54 percent) wanted to "forbid public speeches against democracy." While 46 percent said such speeches should not be forbidden.

The differential impact of these words has been consistent over the years. In the mid 70's nearly half the public would "not allow" a communist to give a speech, but only half that many would "forbid" it. So many more Americans wanted to "not allow" it than to "forbid" it.

You, Webster and I might have difficulty describing what the difference is. But people don't. Those words carry very different connotations. Prohibiting or banning something is much harsher than simply not allowing it.

So it is on questions about a constitutional amendment prohibiting gay marriage.

Let's get the unpleasant fact out of the way - most Americans oppose gay marriage.

But what to do about it? A constitutional amendment says the far right. And what should that amendment say? CBS found 59 percent in favor of a constitutional amendment "that would allow marriage ONLY [sic] between a man and a woman."

Only 35 percent oppose such an amendment.

By contrast, the Annenberg National Election Study showed a plurality opposed to such an amendment. However, Annenberg described the amendment as saying that "NO [sic] state can allow two men to marry each other or two women to marry each other."

When CBS added a clause to their description saying an amendment would allow marriage only between a man and a woman and would also "outlaw" marriage between people of the same sex, a 24 point margin in favor of the proposal shrunk to just 9 points.

One version of the question suggests strong support for an amendment, another moderate opposition, with a third version lying in between. When minor modifications in wording produce significant variation in attitudes, it sometimes signals weakly held views. But those different responses can also indicate differences in meaning attached to words or phrases.

Thus, responses to any one question should never be taken as a perfect measure of the underlying attitude. If something is worth measuring well, the only way to so it is with a variety of question wordings.

Mark S. Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982, including Sen. John Kerry (D-Mass.) this year.



More information about the lbo-talk mailing list