This paper analyzes how the statistical properties of a risk affect th
e attitude of individuals towards accepting another independent risk.
We conduct the analysis for the class of increasing utility functions
having all their derivatives with alternating sign. Such utilities can
be expressed as mixtures of negative exponential functions and they a
re fully described by distribution functions over the set of exponents
. Our analysis exploits the relationship between the distribution func
tions characterizing utilities and the distribution functions characte
rizing risks. In particular, we find sufficient conditions for an addi
tional background risk to either reduce or increase the index of absol
ute risk aversion. (C) 1997 Elsevier Science B.V.