Over on Freakonomics' blog, they have a piece on explaining why accident rates appear to be well correlated to your Sign. They then asked users to come up with why.

Looking at the rankings and the zodiac, it's apparent there's an odd cutoff between Leo and Libra (Leo's being less accident prone, and Libra's being most) when they're only two months apart.

On the Freakonomics blog, user Sage was the first to show that this points to a schoolyear-based cohort of children, and I agree. But Sage doesn't tell how this might become a life-long trait (that's the riddle, Libras are the most car accident-prone for the rest of their lives).

At first, I thought it might have something to do with the fact that kids learn what to do from other kids their age (Dad shows kid that it's safe to swim in the deep end, kid won't do it; kid sees same-aged kid jump in, kid jumps in, too) (I need a citation for this).

But that doesn't show me a really good argument as to why those kids would retain those bad driving habits for a lifetime. So, I finally decided that the better story was that the older children in any cohort are on average physically more developed, have more social/sexual experience, and are the first through any age-based filter that society puts on the group. Because of this, they have been pushing the boundaries of their group ever since they dropped into the cohort. Having been so conditioned, they might remain so for the remainder of their lives.

If the older ones are, as the study suggests, less risk-averse, then we might would expect teenage pregnancy (teen to teen caused), drug use, and other risky behavior to follow the Stars, just as accidents do.

So now I've got a story that makes a testable prediction, but this is all assuming that the original data set wasn't junk. If the data set is good, and the theory is good, then:

  1. If you want a risk-taking child, make sure they are among the oldest in their schoolyear.
  2. If you want a risk-averse child, make sure they are one of the youngest in their class.

But the solution is incomplete, as I can't explain why accident rates would be different than ticket rates. Can you? My brain kinda caves in and just says, "go back to the data, the difference is probably spurious.", which I know is a error in reasoning (if I accepted the data earlier, why stop accepting it now? Just because I'm not smart enough to come up with a link? Sheesh. Some armchair thinker I am. ;)