Simulation-based Value-at-risk for High-dimensional Nonlinear Portfolios

Abstract

[evdate 2016-06-23] IAS Quantitative Finance and Fintech seminar series. IAS Quantitative Finance and Fintech Mini Workshop. Title from slide title.

Abstract: Value-at-risk (VaR) has been a standard risk measure for quantifying market risk since its introduction. While the Delta-normal approach can theoretically be applied to approximate the VaR of portfolios, the accuracy of such an approach, however, significantly diminishes when the portfolios concerned contain derivative positions with nonlinear payoffs. The lack of closed form solutions for these potentially highly correlated derivative prices further complicates the problem. This paper proposes a model-free simulation-based algorithm for VaR evaluation. The proposal leverages cross-sectional information and applies variable selection techniques to simplify the simulation framework, which can be made possible by formulating the problem of interest into a high-dimensional sparse regression problem. Asymptotic properties of this new approach are established which demonstrate the advantages of the additional model selection component in the original least-squares Monte Carlo approach. Numerical results verify the effectiveness mof our approach in comparison with existing practices.

Duration: 32 min.