Maximum variation of total risk (Q1922253)

Motivated by two different scenarios, one of which is related to the question of how extra information affects the distribution of the lifetime pay for a certain life insurance policy, the author defines the partial order \(\preceq\) among random variables by \(Y\preceq X\) iff for every convex \(\varphi\), \(E \varphi (X) \leq E \varphi (Y)\), after which he proves that if \(Z\) is a positive integer-valued random variable and \(\{{\mathcal F}_n\}\) an increasing sequence of \(\sigma\)-fields, then \(Y\preceq {\mathcal E}\), where \(Y = \sum^\infty_{n=1} P(Z = n + 1 \mid {\mathcal F}_n)\) and \({\mathcal E}\) is a standard exponential random variable, a more general continuous-time analog, and shows how the results relate to his scenarios.