The cause of the disparity is party my old bug bear of questions asked as an agree/disagree statement grid. I have railed about these in the past, but in short, while such questions have their specific uses, they are extremely vulnerable to bias.

Let’s start with basic principles. If you are seeking to find out what people’s opinion is on same-sex marriage, a good place to start is a question along the lines of “Do you support or oppose legalising same-sex marriage?”. A possible drawback with this is the issue of civil partnerships, which colloquially many people may think of as gay marriage. Given the issue currently at hand is whether to allow couples of the same sex to marry or just enter a civil partnership, we should really make that clear in the question.

This is basically the approach ICM took in their Sunday Telegraph poll:
--

Since 2005 same-sex couples have been able to enter into civil partnerships. While civil partnerships offer the same legal rights as marriage, same-sex couples are not able to marry. Which of the following best reflects your view?
I would support same-sex marriage
I support civil partnerships, but would oppose samesex marriage
I oppose both civil partnerships and same-sex marriage
Don’t know

They found 43% support for same-sex marriage, 32% support for civil partnership but not gay marriage, 15% opposed to both and 10% sayibng don’t know. Clearly the proportion of people supporting gay marriage is almost identical, but the more nuanced options for those opposed to gay marriage has led to some people who would otherwise have said don’t know giving an opinion.

The other two polls at the extremes were both agree/disagree statement grids. The poll that showed the most support for same-sex marriage, from Populus, asked:

Do you agree or disagree with the following statements?
--
“Marriage should continue to be defined as a life-long exclusive commitment between a man and a woman”

70% of people agreed, with 22% disagreeing. Here we have a statement that doesn’t even mention same-sex marriage, being rather overinterpreted by people for their own ends. Perhaps the logical inference is that anyone agreeing to this statement must be opposed to the laws on marriage being changed so that same sex-couples can marry (though, by the same rather tendentious logic, one could infer that 70% of people must think that divorce should also be made illegal), but in this case, we have other questions in other polls that actually asked directly about gay marriage, so we can be fairly certain that 70% of people are not interpreting the sentence that way.

I would again urge people to be cautious about polling questions asking if people agree/disagree with a loaded statement. It is rarely the best way of asking a question and carries with it risks of bias. If there are a conflicting polls on a subject, do not cherry pick those that suit your own views – take a broad look across all the polls. In this context, when the argument is cast in terms of equal rights a majority support gay marriage, when cast as a minority imposing their views upon the rest a majority are opposed – when asked simply and directly just under half are in support.