@stiltskin - Does anybody really think that is going to make a significant difference either way?
Possibly - let me check some evidence to see if I can be convinced...
This will be a long one!

(Courtesy of University of Adelaide Centre for Automative Safety Research)
Here the speed limit was 60 km/h (our 40 mph).
Excuse the statistics, but that's what we're speaking of when considering that the "average" speed has gone down.
Let's make a few major assumption - let's assume that UK drivers are like Australian drivers (likely) and that the same shape distribution will happen at 30mph limits and 20mph limits in the UK (possibly a bit more shaky ground here!).
The distribution above isn't a "normal distribution", so isn't symmetrical, there are more people drive under the speed limit than over.
The most popular speed (the "mode") is 61 km/h, and (going to the linked page) the average is 59.9 km/h, however 58% of people drive slower than this. 97% of people drive slower than 5 km/h over the average/speed limit.
So 60% of people will be going slower, and most will be clustering around the limit (based on the assumptions I've made above).
That seems pretty convincing, although I'd much prefer to see the distribution at other limits to see if they are similar.
Further down the paper, some scenarios are indeed presented.
A 5km/h speed reduction (similar to the observed average reduction) had a predicted 15% reduction in crashes, and 13% fewer people injured. I note that the TRL figures (from Morningsider) are similar, at 12-18% (scaled to 3 mph reduction), which lends weight to the UK driver=Australian driver argument.
Total compliance with the 60km/h limit was predicted to reduce crashes by 28.6%, and 30% fewer casualties, which would be much better, but that's not the world in which we live! (Hmm, possible evidence for GPS speed limiters?)
A 50km/h limit (our 30mph) with real world compliance would see crashes down by nearly a third, although the percentage injured would be down by only 27% as the speeders disproportionately cause injuries (crash energy varying by the square of impact speed).
So, with only 2 mph reduction in average speeds, I'm quite happy.
Robert