Alex Steer

Better communication through data / about / archive

Freerunning from the law

835 words | ~4 min

This piece from the Independent, about the mooted introduction of parkour lessons in secondary schools, is a textbook example of why you should be careful when using social research data to inform policy. The report says:

According to figures from the Metropolitan Police, when sports projects were run in the borough of Westminster during the 2005 Easter holidays, youth crime dropped by 39 per cent. The following year, the most recent for which figures are available, when parkour was added to the projects, youth crime fell by 69 per cent.

So, was it parkour?

There is a fair body of evidence that providing summer activities for young people, particularly in deprived areas, can cause short-term falls in certain types of youth crime, especially vandalism and various strains of antisocial behaviour. One pilot study, done on an estate in Bristol in 1992, found a 29% drop in overall crime during the time the summer scheme ran, and a 68% drop in vehicle theft. Another, in Runcorn, was cited as the cause of a 57% drop in police callouts for youth disturbance in 1993. (Both examples are from Demos, Turning the Tide. There is little evidence for the long-term impact of schemes like this, though that's a different problem.)

Westminster Council takes youth crime reduction seriously. One of the targets for its 2003-5 Youth Crime Reduction Strategy was to 'develop the Positive Activities enable at risk young people to participate in positive activities during the school holidays'. Importantly, though, when the Strategy was written in April 2003 the activity scheme idea was listed as new. It was part of the Home Office's Positive Futures strategy, launched in 2001, to improve 'social inclusion' using sport and leisure activities. (You know, the kind of thing the Daily Mail loves.)

This already gives us a problem. Because Westminster adopts such a thorough approach to youth crime reduction, especially during the holidays, we can't safely attribute the crime reduction to parkour. (We could, however, attribute increased attendance at Positive Activities to this, though perhaps also to better publicity, word of mouth from last year's kids, etc.)

But there are other problems. I can't find a source for the 38% and 69% drops in youth crime for Westminster in 2005 and 2006. (If anyone knows where I can get them, I'd really like to see.) So I'm going to have to use some proxies. Let's take a look at the number of reported crimes for Westminster over the Easter period in 2006, 2007 and 2008. This is all types of crime, done by criminals of all ages.

March 05: 5399 April 05: 6501 May 05: 6444

March 06: 5399 April 06: 5169 May 06: 5658

March 07: 5831 April 07: 5705 May 07: 5617

March 08: 5264 April 08: 5056 May 08: 5103

This doesn't tell us too much: sometimes it seems to rise over the Easter holidays, sometimes to fall. So let's look instead at the number of reported motor vehicle thefts for 2005 and 2006.

2005: 98 2006: 63

I've chosen motor vehicle theft because it's disproportionately a young people's crime, and seems in other studies to have been heavily improved by the provision of holiday activities. On the face of it, this looks great: a 35.7% reduction in a crime largely committed by youngsters. Clearly something in the Youth Crime Reduction Strategy (though not necessarily parkour) was working.

But let's pull out and look at the figures for the whole period 2000-2008.

2000: 121 2001: 15 2002: 139 2003: 17 2004: 12 2005: 98 2006: 63 2007: 48 2008: 50

Mean: 62.5%

Suddenly the drop doesn't look so impressive. In fact, there's little clear pattern at all. It seems it's a classic case of Regression toward the mean, in a set of numbers with a pretty high distribution (lowest: 12; highest: 139). When Westminster published its Youth Crime Reduction Strategy in April 2003, little wonder it was worried if this statistic is representative: its last set of Easter holiday stats (for April 2002) showed a huge spike of car thefts on the previous year (from 15 in 2001 to 139 in 2002). Yet at the same time the figure was poised to plunge again for 2003: just 17. It's now hovering somewhere comfortably just below the mean.

So has parkour killed off youth crime in Westminster? Not for certain. My proxies tell us a lot less than proper youth crime statistics would, and if I find them I'll report on them here. What I'd like to know, and what I wish the Independent or Westminster or the Met would tell us, is what the figures look like for the last ten years or so. Do they mirror the pattern for vehicle theft? That would make it easier to tell if we're looking at a genuine reduction or just a regression toward the mean.

# Alex Steer (26/01/2009)