Adam Pettersson: Computing a cost-of-capital margin with least-squares Monte Carlo

Abstract:

This thesis evaluates the LSM (least-squares Monte Carlo) method for estimating a certain cost-of-capital margin. We simulate Gaussian Cash flows representing insurance run-offs. Simulating from a Gaussian process is not entirely realistic for the insurance context we consider but allows for explicit computation of the Cost-of-Capital margin derived in Engsner et al. (2017). We use an LSM approach to approximate the Cost-of-Capital margin on the same simulated data. Since the Gaussian nature of the cash flow allows us to calculate the Cost-of-Capital margin analytically, it is possible to assess how well different set-ups on the LSM perform. If an LSM algorithm performs well on Gaussian data, this indicates that the algorithm may perform well on non-Gaussian data. We will also make some connections to the Solvency 2 risk margin.