Graeme wrote:
Interesting how that Sainsbury survey says there were 199K accidents in 2005 and that there are 30M motorists.
My maths gets that to 0.66% - so how does that stack up with 4% being involved in an accident in a 6 month period....

The 199K figure is for injury crashes. The 4% is for all crashes, including 'damage-onlys'. Damage onlys are around ten times more common than injury crashes.
There are now 34million licenced drivers, so 199K injury crashes means 1 caused for every 170 driver-years (assuming each crash is caused by one driver). Since some crashes are caused by non-drivers (pedestrians etc) the true figure is even better.
It's also worth noting that this includes injury crashes caused by highly risky groups (stolen vehicles for example.) The risk value for the median driver is at least five times lower even than this (one injury crash caused for > 850 driver-years).