Mean streets

My mobile phone cheats at backgammon. Odd how it throws itself loads of double-sixes just when it needs them, and gives me a double–five exactly and only when that is the one dice-roll that is completely useless. Funny how the very moment I uncover the ‘four’ point, the machine awards itself a four.

Of course, this is all paranoia; my excuse for being demonstrably the world’s worst backgammon player. The dice rolls are generated at random, but I only remember the few that are spectacularly opportune or desperately inconvenient, dependent on whether they’re the black or the white dice.

If my opponent rolled 2D6 and, after an expected spread of results got three double-sixes, I might sit up and take notice of his lucky streak. I might brandish Harry Potter’s wand and with a “Reductio fortuitus!” cast an incantation to prevent the dice from rolling uncharacteristically high. And the spell works! The next five rolls are not double-six. QED.

The gainsayers, disbelievers and acolytes of James Randi will point out that my eldritch magicks have no effect on the behaviour of the dice. Simply put, rolling three twelves in a row has a 1:46656 chance (0.00214%). Not rolling a double-six for the next five rolls comes at a probability of 86.86%. No magic required.

And this brings me to what continues to bug me.

For some reason, a basic understanding of this ‘regression to mean’ doesn’t apply when so-called ‘safety’ cameras are installed on the roads. For example, following decades of no traffic incidents^[1] at a particular location, there’s suddenly a rash of five KSIs (Killed or Seriously Injured) over, say, a five-year period. A camera is installed, and “Reductio fortuitus!” for the next five years the KSI rate reverts to zero. Clearly it’s the magic box that has achieved all this, and not basic statistics.

We are told that excessive speed is a major contributory factor in traffic incidents. How many? What proportion?

According to the UK’s Transport Research Laboratory’s report TRL323 - “A new system for recording contributory factors in road accidents.” (© TRL 1998, ISSN 0968-4107 and downloadable free of charge from the TRL's website):

• Excessive speed is the overall contributory factor in 7.3% That's an average; actual between 2.8% and 9.2% depending on geographical area.
  
• Excessive speed the ‘definite’ contributory factor in 6.0% of recorded incidents.

These figures came from UK police officers attending crashes and recording the factors involved. The period of data collection was three months in the summer of 1996.Unfortunately other TRL research showed that only 30% of excessive speed actually involved exceeding the posted limit. So 30% x 7.3% = 2.2%. What about the 97.8% of incidents that don’t involve exceeding posted limits? Cameras must surely have zero effect on these!

The trouble is, these figures aren't even approximately close to the Dubai RTA's most recent claim that “Overspeeding is responsible for about 80% of traffic accidents…” The RTA's Director of Traffic goes on to state that “…the number of casualties is continuously on the rise…” something that enormous and increasing numbers of cameras seem powerless to prevent.

Just in case the RTA website is slow loading or broken, 7DAYS picked up the same story.

What has happened? Why has the enormous increase in speed-related KSIs occurred at the same time as the rash of speed enforcement devices? Claiming that cameras cause incidents is tenuous; but the argument that the cameras have a positive effect on road safety is not borne out by the RTA's own statements.

Now, it may be argued that UK and UAE driving standards are different. It may also be pointed out that the TRL report is based on data collected over a short period (of mostly fine dry weather). But from 7.3% to 80%? Asking us to believe the uplift from around 30%, RTA's own figures from 2002 and 2005 is pushing the bounds of believability.

At this point, the Goat checks his portable tosh detector. (The Acme Veritas-O-Mat^TM pat. pending). The needle is waving around in the red zone like some demented instrument in a Thunderbirds episode.

Whom are we to believe? An independent research laboratory, or someone who is employed by the same organisation that earns a tidy income through the use of traffic enforcement and presumably will seek to justify its revenue stream? I know which direction I tend to lean.

It is or course completely true that anyone who never strays above the posted limit isn't going to be photographed and fined. It's also true that “No-one was ever injured in a car crash below the posted limit” is total fiction. There is a lot more to road safety than simply automating the enforcement of speed limits.

[1] ‘accident’ is a misleading term because it implies that no-one’s at fault and ‘traffic deliberate’ doesn’t really work either. Neither are we supposed to call them ‘speed cameras’, even though speed is what they detect. But the cameras don’t positively detect safety. Perhaps ‘danger cameras’ would be a better term. ‘Enforcement’ cameras might be best, but this word has rather unfortunate connotations involving the removal of free will.

]}:-{>