Replicating portfolio approach to capital calculation (Q1691451): Difference between revisions

The authors establish solid mathematical foundations for the replicating portfolio approach to the evaluation of capital. The basic idea of the replicating portfolio, as noted by the authors, is to project the terminal loss of discounted asset-liability cash flows on the space dynamically spanned by a finite number of financial instruments which are easier to value by simulations. The mathematical foundations of the replication portfolio are established using an important and fundamental result in stochastic analysis, namely a chaos expansion, and the approximation of the terminal loss by the replicating portfolio, which is a self-financing portfolio formed by a set of financial instruments, is evaluated by solving a minimization in the norm of the space of square-integrable random variables. One of the main results obtained by the authors is an upper bound on the approximation errors, which is presented in Theorem 3.7. The authors also propose the use of empirical regression to solve the minimization problem. Monte Carlo simulations are employed to compute estimators for the approximations to solvency capitals via conditional tail expectations and value at risk in Section 4. Another main result obtained by the authors is that the estimates from Monte Carlo simulations are asymptotically consistent when the sample sizes in the simulations are getting large. This main result is presented in Theorem 4.2 and Theorem 4.4. Some numerical examples to illustrate the approach are presented in Section 5, which are mainly based on the geometric Brownian motion assumption.