Literally survivorship bias.

True, however, I think survivorship bias tends to imply that a relatively small number of the elements pass through the system, thereby making the data gained from said elements biased. It's like looking at only elements that end up 3 standard deviations higher, and then trying to make remarks on the whole system from there. In this case, the vast majority of children will never ever be harmed walking home from school to the point of death, the mean of the normal distribution can be taken to represent the system accurately. I don't know my logical fallacies enough to name this one, but I think you have this inverted.