Reducing Asset Weights' Volatility by Importance Sampling in Stochastic Credit Portfolio Optimization
by Stephan Tilke of the University of Regensburg
Abstract: The objective of this paper is to study the effect of importance sampling (IS) techniques on stochastic credit portfolio optimization methods. I introduce a framework that leads to a reduction of volatility of resulting optimal portfolio asset weights. Performance of the method is documented in terms of implementation simplicity and accuracy. It is shown that the incorporated methods make solutions more precise given a limited computer performance by means of a reduced size of the initially necessary optimization model. For a presented example variance reduction of risk measures and asset weights by a factor of at least 350 was achieved. I finally outline how results can be mapped into business practice by utilizing readily available software such as RiskMetrics' CreditManager as basis for constructing a portfolio optimization model that is enhanced by means of IS.
Keywords: CVaR, Credit Risk, Stochastic Portfolio Optimization, Importance Sampling, CreditMetrics, CreditManager.