Full downside risk aversion

It is shown that well-behaved notions of greater or less downside risk aversion, via utility transformations, lead not to just one, but two, dual, notions of absolute aversion to downside risk: one, the more evident but weaker condition, requires that the prudence measure be positive, given a positive Arrow-Pratt measure of risk aversion, whereas the other, stronger, but less obvious, condition requires that the prudence measure be greater than three times the corresponding Arrow-Pratt measure. The reason for the appearance of these two extreme conditions, bounding the spectrum of reasonable alternative notions of downside risk aversion, or equivalently of downside risk loving, are explained, and consequences of this divergence in the possible meanings of downside risk aversion are explored.