[dalke]

As a quibble, "not statistically significant" means "cannot tell if there was an effect, either positive or negative, because it's smaller than the noise." The only difference is that one can't use it to argue that it 'theoretically [caused] a (small) net increase in safety'.

You use the phrase "intentional danger". The flip side is that "illusory safety." A 12' lane may feel safer even though it actually isn't. But it feels like you stress the "intentional danger" part when the label is irrelevant - the questions are the number and severity of crashes, of human injury, of overall traffic capacity, etc. How one labels the emotional aspect of the driver's internal state isn't relevant to the outcomes.

Also, as secabeen pointed out, driver vigilance isn't the only concern. To add another one 12' of road surface is simply more expensive to repair, clean, and replace than 10' of surface.

Regarding self-driving cars - sure, but how does that change anything in the next 10-20 years of road design? The underlying factors won't change unless a large percentage of vehicles are self-driving. You might as well argue that enforcing a 10mph speed limit would be safer as well, as another example of a solution which, even though correct should it occur, won't affect anyone's planning now.