Non-parametric moment-based copula estimation: Rodney Wolff (Queensland University of Technology, Brisbane)
Risk measurement can often depend on understanding joint distributions between two or more variables, or risk factors. One approach to this is to make use of copulae. The standard approach to copula estimation is to assume a functional form and to estimate associated parameters. We employ asymptotic expansions of distributions to show how one might use, in a very simple way, moments and cross-moments to estimate a copula non-parametrically. Illustrations using index returns and associated implied volatilities will be given to demonstrate the method, and to identify its limitations. Technical details will be given where required, but will be explained in an intuitive fashion.
Joint work with Kohei Marumo (Bank of Japan; QUT)
For more information please contact Dr. M. Anufriev