By Chris Dillow

These remarks by Jack Straw irritate me:

"There is a specific problem which involves Pakistani heritage men... who target vulnerable young white girls. We need to get the Pakistani community to think much more clearly about why this is going on and to be more open about the problems that are leading to a number of Pakistani heritage men thinking it is OK to target white girls in this way."

My problem is that this sort of claim bundles up trivially true statements with contentious ones. It’s trivially true that sex crimes, however rare, are a problem. And it’s trivially true that some Pakistani men are sex criminals, just as some whites are. The question is: why bring ethnicity into it? Are men of Pakistani heritage more likely to commit sex crime than others?

There is a reasonable motive for asking this question. Maybe Islam’s different attitudes towards sexuality and women create a different predisposition among Pakistani men towards sex crimes. This is a reasonable hypothesis, to be tested against the facts. But there’s also an unreasonable motive - the outgroup homogeneity bias generates an unreasonable double standard: if a white man commits a rape, he’s just a rapist but if a Pakistani does so, he’s a Pakistani rapist. You wouldn’t ask the “white community” to look into itself if a white guy commits a sex crime, so why ask the “Pakistani community” to do so if a Pakistani does so?

And herein lies my irritation. Straw gives us no statistics to justify his claim.

Those that do exist seem to undermine his claim.

Table 5.4b of this pdf shows that, in the latest year for which we have data, Lancashire police arrested 627 people for sexual offences. 0.3% of these were Pakistanis. That’s two people. 85.5% were white British. In Lancashire, there are 1,296,900 white Brits and 45,000 Pakistanis. This means that 4.163 per 10,000 white Brits were arrested for a sex crime, compared to 0.44 Pakistanis. If you’re a journalist, you might say that the chances of being arrested for a sex crime are nine times greater if you’re white than Pakistani. If you’re a statistician, you might say they are 0.037 percentage points greater.

Now, you can quibble with these figures. Arrests could be a biased measure of the ethnic prevalence of crime: upwardly so if the police are racist, downwardly so if political correctness leads the police to soft-pedal investigations into Pakistanis’ crimes. But if Mr Straw thinks the police are failing to investigate crime properly, he should say so.

I don’t say this to make any point about ethnicity. I do so more to ask for better standards of political discourse. If you think there’s a problem of crime amongst a particular ethnic group, give us hard evidence of this; feel free to offer me some different figures than those I‘ve cited.

This matters because when we are discussing low-probability events - crime, risk, whatever - we are prone to all sorts of cognitive biases. The way to correct such biases is to use statistics. In not doing this, Mr Straw is inviting his audience to draw some inferences which might not be warranted.

Originally published at Stumbling and Mumbling