Random Risk Appetite

Abstract

There is a burgeoning literature on the randomness of the coefficient of relative risk aversion (CRRA). This paper is in line with such a research agenda. Modelling risk aversion, or its converse, risk appetite, as a random variable violates one of the fundamental principles of economics, in general, and of the behavior under risk in particular, and which is constant preferences. This paper argues otherwise. Both conditional and unconditional tests are carried out to identify the CRRA. A battery of econometric procedures is attempted. The paper postulates that the CRRA follows a normal distribution, with the first two statistical moments derived from the empirical results. The CRRA is found to follow a normal distribution with mean 2.57, and with a standard error of 0.454. Surprisingly, the 95% confidence interval does not include a CRRA of +1, or log utility. However the richness of the approach compensates for this caveat.