16 year olds get better at driving.

They also get less likely to commit crime, but that’s not how we gauge risk. We don’t generally say “that teenager’s crime risk is going down so they are less risky than that geriatric whose crime risk is fairly constant.” Risk probability is usually a hazard rate integral across an interval.

Over a long enough interval, that reduction in risk would be important. So what is the appropriate time interval for these risk assessments?