Unspinning Evangelical Marriage Poll Numbers

Andrew Sullivan is surprised and encouraged by a poll showing that white evangelicals are human beings with a variety of concerns, which they approach with varying emphasis. Who'd have thought such a thing could be possible?

While I must admit to some reluctance to temper Sullivan's revelation, it's important to note that he gets the numbers wrong. Some of the blame belongs with his source and its source, both of whose contributions are probably some combination of error and spin. Here's Sullivan's first claim of fact:

52 percent of evangelicals said they preferred the matter to be handled by the states.

Here's how Frank Langfitt of the Sun-Sentinel (which I'm pretty sure is Sullivan's source) put it:

52 percent would prefer to rely on state laws to prevent gays from marrying rather than altering the U.S. Constitution.

That language comes directly from the PBS Religion & Ethics Newsweekly summary of its survey. But they're all wrong. Here's the actual question that respondents answered (see page 12 on this PDF):

Should the U.S. Constitution be amended to ban gay marriage, or is it enough to prohibit gay marriage by law without changing the Constitution?

The first thing to note is that the number is actually 52% of the 85% who oppose gay marriage. More importantly, the summary writer inserted the concept of state law and introduced the notion of preference, and then Sullivan replaced the emphasis on prevention with generic "handled by the states." The question, itself, however, is one of effectiveness — whether regular ol' laws, federal or state, would be sufficient. The survey doesn't explore whether proven inadequacy of statutes would switch support toward an amendment. Mention of the judiciary might have yielded higher support for an amendment, for example.

Here's Sullivan's second claim of fact:

Moreover 48 percent of evangelicals said that support for marriage rights for gays would not disqualify a candidate from their vote.

He actually got that backwards, as checking his source shows:

only 48 percent of white evangelicals said a candidate's support for gay marriage would disqualify him from receiving their votes.

The "only" is a bit of mild editorializing on Langfitt's part, not included in the PBS summary. Still, PBS assists him by failing to mention that the 48% compares with only 38% who would still vote for a candidate who disagreed on the issue.

The number mix-up is probably an honest mistake on Sullivan's part. However, to the degree that he strengthens the language with his rhetorical boilerplate ("support for marriage rights for gays"), he highlights two important omissions that begin with the summary.

First, another 7% said it "depends." The question is sufficiently open regarding both the degree of disagreement and the areas of agreement that some folks (myself included) would leave room for an election offering the choice between, say, a candidate who would abandon the war on terror and issue an executive order instituting gay marriage and one who would be strong against terrorism and offered some mild, qualified support for gay marriage. Seen in such a light, the 48% is astoundingly high.

Second, if I'm reading the survey data correctly, this question was asked of the total population. In other words, the 38% who would vote for a candidate with whom they disagreed about gay marriage might include as much as 10% who would — despite disagreement — be voting for a candidate who opposes gay marriage.

Thus do numbers transform in meaning and significance. Polls don't offer the most concrete of data, in the first place, but as the layers of spin and misreading pile on, they shift from cautiously useful, to useless, to detrimental.