Jared Diamond writes in Collapse (which I've been reading and think is good):

consider a narrow river valley below a high dam, such that if the dam burst, the resulting flood of water would drown people for a considerable distance downstream. When attitude pollsters ask people downstream of the dam how concerned they are about the dam's bursting, it's not surprising that fear of a dam burst is lowest far downstream, and increases among residents increasingly close to the dam. Surprisingly, though, after you get just a few miles below the dam, where fear of the dam's breaking is found to be highest, concern then falls off to zero as you approach closer to the dam! That is, the people living immediately under the dam, the ones most certain to be drowned in a dam burst, profess unconcern. That's because of psychological denial: the only way of preserving one's sanity while looking up every day at the dam is to deny the possiblity that it could burst. (p436)

Reading this, I wonder: how do we know this is not an effect of the people most concerned about living right under the dam moving away?