National Geographic has an excellent article on teenage risk-taking and adolescent brain development.
It goes some way to explaining both the dangerous mistakes that typically peak in the late teens and, I like to think, the bad fashion sense which seems to follow a similar pattern.
Importantly, the piece goes beyond the usually ‘well the frontal lobes are still developing, aren’t they?’ explanation that gets wheeled out whenever teen neuroscience is discussed and hits on some of the gritty details.
Are these kids just being stupid? That’s the conventional explanation: They’re not thinking, or by the work-in-progress model, their puny developing brains fail them.
Yet these explanations don’t hold up. As Laurence Steinberg, a developmental psychologist specializing in adolescence at Temple University, points out, even 14- to 17-year-olds—the biggest risk takers—use the same basic cognitive strategies that adults do, and they usually reason their way through problems just as well as adults. Contrary to popular belief, they also fully recognize they’re mortal. And, like adults, says Steinberg, “teens actually overestimate risk.”
So if teens think as well as adults do and recognize risk just as well, why do they take more chances? Here, as elsewhere, the problem lies less in what teens lack compared with adults than in what they have more of. Teens take more risks not because they don’t understand the dangers but because they weigh risk versus reward differently: In situations where risk can get them something they want, they value the reward more heavily than adults do.
Probably one of the most comprehensive introductions to teen risk you’ll read in a good while.
Link to National Gerographic on Teenage Brains.
Mind Hacks 
funny you posted this today. I was just thinking about teenage risk taking yesterday.
Looking at the risk/reward (expected loss vs. expected return) trade off, you mentioned how teenagers might evaluate reward differently.
There might also be a difference in how teenagers and older adults gauge expected loss. To give an example, let’s consider a particular extreme risky behaviour that say has a 10% chance of death (r=0.10).
Now, there are different ways you can measure the expected loss. The first thing that came to my mind was to count ‘all the years to come’ that might be lost and thus never experienced. Looking at it this way, and with an over-simplified assumption of valuing all years the same, the expected loss for a 15 year old with a life expectancy of 80 years, will be (80-15)*0.10 = 6.5 years, while the expected loss for a 50 year old adult with the same life expectancy would be (80-50)*0.10 = 3 yrs. Since 6.5 > 2 * 3, the teenager would have to value the possible reward more than twice as much as the adult in order to justify the risk. That might very well be the case, but I suspect there could also be another factor at play here.
I suspect, the way one evaluates the risk of death, is partly based on the loss of all those years to come, but it is also partially based on the loss of memory of all the years past. Looking at it this way, and again with much simplifying assumptions,
the expected loss for the teenager and the adult might be written as (80-15)*0.10 + 15*f and (80-50)*0.10 + 50*f respectively, where f is the factor for measuring the value of the years experienced. (f==r and the expected loss would be the same for the teenager and the adult, f>r and the expected loss would be higher for the adult).
I was tempted to interpret this as the sum of the individual loss and the societal loss, i.e. interpreting loss of experience as a communal loss. But I am not so sure about it any more. In fact I now think the second factor might be even more personal than the first, kind of like being ‘vested’ in one’s life!