I think you'd really also need to normalize by age of the driver. I suspect that younger (under 21, say) drivers are more dangerous in both countries, but that in places where it's expensive and time consuming the licensed drivers will tend to be older.

I don't see why you'd normalize for that. I don't doubt that age is a factor, but surely the effect due to age should actually be counted. For example, different safety standards in cars will result in different fatality rates. That doesn't mean you try to eliminate that from the picture by normalizing for it, that means that the country with better safety standards is actually safer. If a country's roads are safer because they discourage or prevent younger drivers from driving, should not that additional safety count? Taken to an extreme: if a country allows 8-year-olds to drive cars and has a massive road fatality rate because of it, that shouldn't be normalized away when comparing with a country where you can't get a license until 18, right?

So my supposition is that although you're allowed to drive at 18 in France, the burden they place on licensing reduces the number of 18 year olds who actually do so, which has the effect of artificially aging their population of drivers. If you want to know how effective the training is, then you might need to normalize for age to eliminate that effect. If the discussion were about differences in how old you need to be to drive, then I'd agree with you, but the article is about the burdensome regulation around driver training.

If the regulations disproportionately discourage more dangerous demographics from driving, shouldn't that count as a safety improvement just as much as prohibiting them from driving altogether? Burdensome regulations are keeping a large number of dangerous drivers off the road. Shouldn't that be considered a win? (Not necessarily one that outweighs the costs, but in isolation that's a good thing, no?)

I'm not saying it isn't a win, necessarily. I'm saying we don't know if the training is what's causing the win, or if it's just the different demographic. Maybe you could get the exact same effect by either raising the driving age to 21, or keeping the driving age the same but otherwise (through fees and other burdens) discouraging people under 21 from driving. If that were the case (and I have no idea if it is or not), then you'd be unnecessarily costing people time and money, to no good effect.

Fair points, but on the other hand, part of the purpose of training can be to weed out people who are unsuitable for doing whatever you're training for. Maybe there are better ways to do it, but even if rigorous training just weeds out the reckless people and doesn't help the others, I'd still call that an advantage of rigorous training. Of course, it's reasonable to wonder if it helps the others too.

Yep, makes sense. Would be great to know if we could skip the test for people over age X, but maybe just making it expensive for people under that age is a good idea.