Maximum Variation of Total Risk

Let Z>0 be a random time. The total risk of discovering Z in the next time interval (t,t+dt) is never more variable than an exponential of mean one, which is achieved when the information up to time t is sigma(Z wedge t).