Surveying sexuality

A bit off topic for this blog, but the recent ONS report on sexual identity (pdf) raises some interesting questions. As pointed out here the estimate of the number of gay/lesbian/bisexual people produced is way lower than that which came from the last National Survey of Sexual Health and Lifestyles (1.6 per cent of the population, compared with about 6 per cent).

Now, the Integrated Household Survey (IHS), from which the new numbers come, is a truly gigantic survey, with around 450,000 respondents. This means we can place a high degree of certainty around the estimates it comes up with. But is it necessarily the case that the new figure must be 'the answer'? Actually, I'd be very surprised if it was. For one thing, the variation by age is highly suspicious - why should there be as much variation by age as the IHS data seem to suggest: surely the proportions of gay and lesbian people should be more or less constant across age? Ditto the class distribution - why should gay people be over-represented in the professional social classes?

One reason is surely that what is captured by surveys such as the IHS is precisely sexual identity, not sexual preference. This was what the survey was designed to do. The 1.6 per cent of people who report being gay/lesbian/bisexual are those who to whatever extent see their orientation as a component of their identity and, if you like, acknowledge this. Those who don't see being gay as a component of their identity (and who may even be 'in denial') will not, even if their sexual orientation is homosexual. The variation by age and class becomes much easier to understand if you look at things from this angle.

So the survey has not produced an estimate of the number of 'gay people' at all, but rather an estimate of the number of people with a gay/lesbian/bi identity. It is therefore almost by definition an underestimate of the 'true' number (as the 1.5 million profiles on Gaydar, reported in the Guardian piece, would seem to suggest). And that's even before we start to think about response rates, sample design, etc.

This should not be a problem as long as we are all using, and talking about, the same set of definitions, But, of course, we are not, so all of a sudden the old estimates are myths and da gayz shouldn't have as much money spent on them as they do.

This is a really good example of the pitfalls inherent in collecting statistics of this type, and the inherently political nature of much data gathering and use. The extent of these issues is such that sometimes it seems almost counterproductive to collect data of this kind, as it will inevitably be misused. But only almost, and, I think, we should no allow wilful ignorance and journalistic bad-faith put us off!