Mean-variance hedging via stochastic control and BSDEs for general semimartingales (Q1931322)

The goal of mean-variance hedging is to minimize the mean squared error between a given pay-off and the final wealth of a self-financing strategy trading in the underlying assets. There are two main approaches to deal with this problem; one of them uses martingale theory and projection arguments, whereas the other views the task as a linear-quadratic stochastic control problem and uses backward stochastic differential equations (BSDEs) to describe the solution. Combining the tools from both areas enables the authors to improve earlier work in two directions. On the one hand, the solution can be described more explicitly than by the martingale and projection method. On the other hand, it is possible to work in a general semimartingale model without restriction to particular set-ups (like Itō processes or Lévy settings). It is shown that the value process of the stochastic control problem associated to mean-variance hedging has a quadratic structure. The three coefficient processes can be described by semimartingale BSDEs, and the authors show how to obtain the optimal strategy from the latter. In Section 4, equivalent alternative versions for the BSDEs are provided which are more convenient to work with in some examples with jumps. Section 5 illustrates some applications of the results obtained, and both sections discuss in more detail connections to the existing literature.