Fulfilling solvency II regulations using high performance computing

Abstract

Throughout Europe, Solvency II Regulations are changing the way in which companies involved in the provision of financial services must assess their solvency. Historically, solvency has been assessed using a single 'best estimate' set of assumptions. Solvency II requires that a Monte Carlo approach is used to find a 1-in-200 worst-case scenario; this demands computing power which is outside the realms of anything currently available in the industry. In this paper, we demonstrate that the new regulations can be met by moving away from the currently-used actuarial valuation software packages, and running well-performing ab initio code in an HPC environment. Our implementation uses a combination of algorithmic improvements, serial optimisations and large scale parallelism which allows a complete assessment calculation on a representative portfolio of annuities in well under one hour. This work brings the Monte Carlo simulations within the bounds of practicality.