This is a thoughtful paper evaluating the effect of speed cameras on the beltway around Barcelona. Bottom line - 30% reduction in accidents, but note how difficult this type of research is. The basic comparison is before and after the cameras, of course, but there is plenty of other stuff going on which might affect the accident rate. So you have to model the trend before with an equation which seems to predict what is going on, and then show that the equation needs a new element to describe what happened afterwards. This extra element is the effect of the cameras. But the maths are really difficult - we just have to rely on the Journal's referees to make sure they got it right. (In this case there seem to be a load of sin and cos in the equation presumably because seasonal ups and downs?)
The authors also point out that in the rush hour no-one can get anywhere near the speed limit anyway so you wouldn't expect to see much effect. The data are police data, so you have to assume it is reliable and didn't change when the cameras came in. In the UK there is evidence that the police are recording fewer accidents even though health data show no change in hospital admissions or deaths.
The Barcelona study had an added strength - they also looked at accidents on other main roads in Barcelona. These other roads showed no change in accidents.
They weren't able to say directly that the cameras saved lives because only one or two deaths per annum on the beltway, but it seems a reasonable logical conclusion.