Wednesday, 1 June 2011

Boris bungles bus crime stats

"There are lies, dammed lies and statistics," Benjamin Disraeli is supposed to have said. Perhaps years from now the citation will be modified - "there are lies, dammed, lies, statistics and Boris Johnson," for the Mayor of London has once again demonstrated a poor grasp* of numbers relating to crime.

In his campaign literature for next year's Mayoral election, Boris claims that under his watch crime on buses has fallen 30% - an impressive statistic, but one that doesn't stand up to all that much scrutiny.

According to figures from Transport for London, analysed by Liberal Democrat Chair of the London Assembly Transport Committee Caroline Pidgeon, any drop in the aggregate number of crimes on buses masks significant variation between boroughs - with 11 out of 32 London Boroughs reporting an increase in bus crimes.

Many of these boroughs experienced significant increases - bus crime in Southwark went up by 5.2%, in Merton by 9.4%, Tower Hamlets by 11.5% and Barking and Dagenham by a staggering 17.8%. You can take a look at the full statistics here, which include borough-by-borough and crime-type breakdowns as well as comparisons with previous years.

These stats show, of course, that in many places bus crime has dropped significantly - by over 20% year-on-year in some cases - but that it remains a serious and increasing problem for many passengers.

Boris has been in trouble before for his use of crime statistics - earlier in the year the the UK Statistical Authority felt moved to write to the Mayor, chastising him for making political use of crime statistics and warning that he risks eroding 'public trust in Government statistics.' Boris was asked by the UKSA's chairman Sir Michael Scholar to comply with their Code of Practice (he has refused) which says that official statistics should be presented without spin or political colour, and that Boris' press release trumpeting the drop in crime "was poor practice, and was damaging to public trust in the statistics produced by Transport for London."

There's another way in which Boris' use of the bus crime stats are a little naughty - as Guardian blogger Dave Hill points out, there had been a long-term trend towards a reduction in bus crime since before Boris took over the Mayoralty from Ken Livingstone, and that Boris' time in power is likely to have had little effect on that trend either way. For Boris to take credit for the fall in bus crime - where there has even been a fall - is stretching things somewhat.

Why am I blogging about the use of bus crime stats - why should this matter?

The public rightly pays a lot of attention to crime statistics - rightly or wrongly they give us a handle on how safe our neighbourhoods are, how likely we are to fall victim to a crime, how we decide to go home at night in some cases. Although we might not have a crime ready-reckoner to hand every time we board the Number 55, publication and reporting of crime statistics does impact on our behaviour, our psyche - it would be remiss of any politician or journalist to claim that our buses (or streets) were safer than they are just as it is for them to stoke our fear of crime by claiming the opposite.

For a geek like me (not a stats geek admittedly, I hardly know the difference between a Bonferoni post-hoc correction and a Mann Witney U...), the use of science and statistics in politics matters because I'd like to see public policy based on evidence - reliable, honestly presented evidence, not stats that are spun to make a particular case. I accept that politics and governance aren't as straightforward as mechanically implementing that which is statistically significant, but if we can't trust the numbers we read, whether or bus crime, drug use or the NHS, we can't evaluate whether public policy is based on anything other than dogma.

So this is why we need more transparent, honest reporting of crime stats from Boris and other politicians - because as things stand, all we have are lies, dammed lies and manipulated statistics.

No comments: