Published online by Cambridge University Press: 28 September 2018
We propose a consistent and computationally efficient 2-step methodology for the estimation of multidimensional non-Gaussian asset models built using Lévy processes. The proposed framework allows for dependence between assets and different tail behaviors and jump structures for each asset. Our procedure can be applied to portfolios with a large number of assets because it is immune to estimation dimensionality problems. Simulations show good finite sample properties and significant efficiency gains. This method is especially relevant for risk management purposes such as, for example, the computation of portfolio Value at Risk and intra-horizon Value at Risk, as we show in detail in an empirical illustration.
We thank Gurdip Bakshi (the referee), Cecilia Mancini, Roberto Renò, Lorenzo Trapani, and David Veredas for useful comments and suggestions. Perez acknowledges the financial support received by the Arthur Wesley Downe Professorship of Finance and the Social Sciences and Humanities Research Council of Canada. Fusai acknowledges the financial support received by Università del Piemonte Orientale. This article has been presented at the Arizona State University Economics Reunion Conference. A previous version of this article was circulated with the title “Multivariate Lévy Models by Linear Combination: Estimation” and has been presented at the 2014 International Conference of the Financial Engineering and Banking Society (FEBS). We thank all the participants for their helpful feedback. The usual caveat applies.