How the government uses dirty data to legislate morality

So what's a standard deviation?

Convenient categories

The problem with evidence-based policy in this immensely controversial area start here, with basic categorisation. They are then compounded by selectivity within the research, with government in particular keen to select measures, sample, reports, and effect to suit their basic agenda.

According to Kinsey, around one-third of US men had at least one homosexual experience in their lifetime. Does that mean that one-third of Americans are gay? Probably not.

What then is the meaning of "being gay"? A spat is currently developing between Consenting Adult Action Network (CAAN) and the Equality and Human Rights Commission (EHRC) over the nature of "sexual orientation". According to the Equality Act 2006 (pdf), the EHRC is concerned with inequalities which affect a variety of groups – including groups defined by their "sexual orientation."

Clair Lewis, National Convenor for CAAN, asked the EHRC why they view homosexuality as an orientation – but not, say, sado-masochism. For the EHRC, Director of Public Policy, Andrea Murray, argues that the EHRC is limited to what the law allows it to consider, and that by definition, “orientation” excludes bdsm: CAAN remain unconvinced and have indicated that they consider this to be ducking the question, that the only answer they have received to date is a hint from their Chair, Trevor Phillips that the difference lies in whether someone is "born that way” or makes a conscious choice.

Whilst such a view may apply to gender, it is questionable whether the nature/nurture argument can ever be so clear in respect of behaviour. In general, genetic causation is not a fixed effect, but varies according to the levels of social variability to which an individual is subject. In a highly regulated society, in which social effects have been minimised, the proportion of genetic causality behind any behaviour will increase - and vice-versa, in a society where difference is tolerated, the genetic effect will diminish.

This issue of categorisation shades into questions of measurement selection in the much more serious area of rape. It is a commonplace of news reporting that the number of rapes has gone up over the last few decades from 1,200 reported in 1980, to over 12,000 in 2004-5, whilst the conviction rate has dropped from a relatively re-assuring 38 per cent to a mere 5.5 per cent in the same period.

The problem with such Shock!-Horror! comparisons is that a complex picture is reduced to a few very basic measures. Conviction rates (here, the ratio between crimes reported and convictions obtained) has plummeted for almost all categories of crime in the same period - and to similar degree. The current conviction rate for rape is certainly not good – but it is not out of line with that for many other offences, such as wounding or burglary.

Public perceptions of what constitutes rape – and so the threshold for reporting an incident - have almost certainly shifted over the period, whilst the basis for even such basic measures as recorded crime has changed.

Sponsored: Network DDoS protection

Next page: Poppycock problem