[kazazes]

I took this survey in 2011. Teachers were not allowed to be present while we took it. There was no moral pressure to be honest beyond "answer to the best of your abilities" so as not to alienate the group. The entire room openly read the questions and people answered with varying degrees of absurdity. Whatever the results of that room were, they were not representative of the group's risk taking behavior.

Take this anecdote as you will, but it will take me a lifetime to be convinced that a faceless government form can poll the group of cards closest to a teen's chest with any aggregate accuracy.

[jldugger]

Well, it may be that the numbers are under / overreported, but the test could still be valid, in the sense that shifts in reported behavior are correlated with shifts in actual behavior. In other words, if you assume today's kids are no more likely to lie on a survey than in the past, you can get a sense of the direction of trends, if not the absolute values.