Benchmark values for higher order coefficients of relative risk aversion (Q2443952)

Let \(X\) and \(Y\) be two random variables valued in some interval \([z_1, z_2]\) with distribution functions \(F\) and \(G\). Let \( F_1(z)=F(z),\;G_1(z)=G(z) \), and \[ F_{k+1}(z)=\int\limits_{z_1}^{z}F_k(t)\text{d}t,\quad G_{k+1}(z)=\int\limits_{z_1}^{z}G_k(t)\text{d}t \] for \(z\in [z_1,z_2]\) and \(k=1,2,\dots\). A random variable \(X\) is said to be smaller than \(Y\) via \(s\)-th order stochastic dominance (\(X\preceq_{\;s} Y\)) if \(G_s(z)\leq F_s(z)\) for all \(z\), and if \(G_k(z_2)\leq F_k(z_2)\) for \(k=1,2,\dots ,s-1\). Outcomes of the product \( ((1-I_1)X+I_1Y)((1-I_2)\,a+I_2\,b)\) are considered by means of utility theory. Here (\(I_1,I_2\)) is a couple of binary random variables with covariance \(\varrho\), vector (\(I_1,I_2\)) is independent of \((X, Y)\), and \(a,b\) are two positive constants. The following assertion is typical. Let \(w\geq 0\) be a given initial wealth level; \(u\) be a utility function such that \((-1)^{k+1}u^{(k)}\geq 0\) for \(k=1,2,\ldots\); \(X\) and \(Y\) be two non-negative independent random variables such that \(X\preceq_{\;s} Y\); \(a<b\) be two positive constants. If, in addition, \(-x\, u^{(k+1)}(w+x)/u^{(k)}(w+x)\geq k\) for \(k=1,2,\ldots, s\), then the expected utility of outcome \(w+((1-I_1)X+I_1Y)((1-I_2)\,a+I_2\,b)\) decreases as parameter \(\varrho\) increases. The applicability of the obtained results for solving some economics problems is shown.