Ordering risk : expected utility theory versus Yaari's dual theory of risk